Post on 19-Jul-2020
ABSTRACT
Title of thesis: THERMAL CHARACTERIZATIONOF FIREBRAND PILES
Raquel Hakes, Masters of Science, 2017
Thesis directed by: Professor Michael J. GollnerDepartment of Fire Protection Engineering
Over the past several decades, the severity of wildland-urban interface (WUI)
fires has increased drastically, resulting in thousands of structures lost globally each
year. The cause of the majority of structure losses is ignition via firebrands, small
pieces of burning material which are generated from burning vegetation and struc-
tures. In this thesis, a methodology for studying the heating to recipient fuels
by firebrands is developed. Small-scale experiments designed to capture heating
from firebrand piles and the process of ignition were conducted using laboratory-
fabricated cylindrical wooden firebrands. The methodology compares two heat flux
measurement methods. Experimental results compare the effects of varying fire-
brand diameter, pile mass, and wind speed. An ignition condition is described using
temperature and heat flux.
THERMAL CHARACTERIZATION OF FIREBRAND PILES
by
Raquel Sara Pilar Hakes
Thesis submitted to the Faculty of the Graduate School of theUniversity of Maryland, College Park in partial fulfillment
of the requirements for the degree ofMasters of Science
2017
Advisory Committee:Professor Michael J. Gollner, Chair/AdvisorProfessor Stanislav I. StoliarovDr. Jiann C. Yang
© Copyright byRaquel Sara Pilar Hakes
2017
Acknowledgments
I would like to thank the many, many people who have helped me along the
way. A big thank you particularly to my advisor, Professor Michael Gollner, for the
opportunity to work with him over the past several years. This project in particular
gave me the opportunity to grow a lot as a researcher. Thank you very much for
your time and advice throughout this project.
I would also like to thank my committee members, Dr. Jiann Yang and Pro-
fessor Stanislav Stoliarov. In addition to taking the time to read this thesis and
provide their thoughts and feedback, they both provided support to the project as
I was conducting it. Thank you to Dr. Yang for providing ideas and suggesting
papers relevant to the problem throughout the project. Thank you to Dr. Stoliarov
for allowing me the use of his cone calorimeter and the help of his students.
I was very fortunate to have several other students at the University of Mary-
land to help me throughout this project. I am very grateful to Hamed Salehizadeh for
the time he spent running many, many experiments. I am also grateful to Matthew
Weston-Dawkes, for the time he spent setting up and helping to run experiments.
Thank you to Erin Griffith, Alicea Fitzpatrick, and Seth Lattner for jumping onto
the project towards the end to help finish up calibrations. Thank you to the oth-
ers that contributed time helping on experiments and in the lab: Evan Sluder and
Nicole Welch.
I was particularly lucky to work with Sara Caton as we both began Masters on
firebrand projects. I am grateful for her support and companionship in the lab and
ii
the office. In general, I was fortunate to work in a lab group of fun and interesting
people. I had the privilege of working in a department full of many wonderful
colleagues. Thank you to the many friends who ate lunch with me, talked through
research thoughts, and in general made the office a great place to work.
I would like to acknowledge the financial support, advice, and encouragement
I received from the WUI Group at NIST.
I would like to acknowledge Tim Holy, who made the MATLAB function
distinguishablecolors.m, which I used for plots included in this thesis.
Finally, I would like to thank my family and my fiance, Will. All of you
have been my biggest supporters and believed in me completely. Thank you for
encouraging me to continue with the Master’s when I was uncertain about concussion
recovery. I am always grateful for your love.
There are many others who provided me with advice, support, encouragement,
and feedback. Thank you to all of you.
iii
Table of Contents
List of Tables vi
List of Figures vii
List of Abbreviations xiii
1 Introduction 11.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41.2 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2 Literature Review 82.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82.2 Ignition: Flaming and Smoldering . . . . . . . . . . . . . . . . . . . . 9
2.2.1 Theories of Ignition . . . . . . . . . . . . . . . . . . . . . . . . 112.3 Experimental Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3.1 Firebrand Production . . . . . . . . . . . . . . . . . . . . . . . 132.3.2 Firebrand Ignition of Vegetative Fuels . . . . . . . . . . . . . 142.3.3 Firebrand Ignition of Solid Fuels . . . . . . . . . . . . . . . . . 16
3 Experimental Methodology 183.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183.2 Calibration Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203.3 Heat Flux Tests With Firebrand Piles . . . . . . . . . . . . . . . . . . 24
3.3.1 Experimental Design . . . . . . . . . . . . . . . . . . . . . . . 253.3.2 Wood Properties and Firebrand Production . . . . . . . . . . 28
3.4 Experimental Procedures . . . . . . . . . . . . . . . . . . . . . . . . . 313.4.1 Firebrand Pile Tests Using Thin-Skin Calorimeters . . . . . . 323.4.2 Firebrand Pile Tests Using Water-Cooled Heat Flux Gauge . . 33
3.5 Tests Under Forced Flow Conditions . . . . . . . . . . . . . . . . . . 353.5.1 Ignition Test Set-up . . . . . . . . . . . . . . . . . . . . . . . 353.5.2 Thin-Skin Calorimeter Forced Flow Experiments . . . . . . . . 38
iv
4 Measurement Techniques and Calibration Results 404.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 404.2 Heat Flux: Water-Cooled Heat Flux Gauge . . . . . . . . . . . . . . . 40
4.2.1 Results of Gauge Size Comparison . . . . . . . . . . . . . . . . 414.3 Heat Flux: Thin-Skin Calorimeters . . . . . . . . . . . . . . . . . . . 42
4.3.1 Fabrication . . . . . . . . . . . . . . . . . . . . . . . . . . . . 434.3.2 Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 454.3.3 Theoretical Formulation . . . . . . . . . . . . . . . . . . . . . 474.3.4 Results of Calibration . . . . . . . . . . . . . . . . . . . . . . 504.3.5 Application of Calibration to Experiments . . . . . . . . . . . 52
5 Results of Firebrand Pile Tests 595.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 595.2 Visual Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 595.3 Thermocouple and Thin-Skin Calorimeter Temperature Results . . . 615.4 Heat Flux Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.4.1 Components of Heat Transfer . . . . . . . . . . . . . . . . . . 655.4.2 Spatial Distribution of Heat Flux . . . . . . . . . . . . . . . . 67
5.5 A Comparison of Heating From Piles . . . . . . . . . . . . . . . . . . 735.5.1 Effect of Brand Diameter . . . . . . . . . . . . . . . . . . . . . 735.5.2 Effect of Pile Mass . . . . . . . . . . . . . . . . . . . . . . . . 775.5.3 Peak Heat Flux and Net Heating . . . . . . . . . . . . . . . . 80
6 Results of Ignition Tests 896.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 896.2 Visual Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
6.2.1 Description of the Ignition Process . . . . . . . . . . . . . . . 926.3 Thermal Characteristics at Ignition . . . . . . . . . . . . . . . . . . . 94
6.3.1 Temperatures . . . . . . . . . . . . . . . . . . . . . . . . . . . 946.3.2 Heat Flux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 976.3.3 Spatial Distribution of Heat Flux . . . . . . . . . . . . . . . . 99
7 Discussion 1037.1 Firebrand Pile Heat Flux Tests . . . . . . . . . . . . . . . . . . . . . 1037.2 Ignition Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
8 Conclusions 1088.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1088.2 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1088.3 Recommendations for Future Work . . . . . . . . . . . . . . . . . . . 110A.1 Raw Thin-Skin Calorimeter Results . . . . . . . . . . . . . . . . . . . 113A.2 Raw Water-Cooled Heat Flux Gauge Results . . . . . . . . . . . . . . 120A.3 Gauge Comparison Plots . . . . . . . . . . . . . . . . . . . . . . . . . 125
Bibliography 132
v
List of Tables
3.1 The test matrix. For each type of test, experiments were conductedusing the diameter and masses of firebrands indicated. Note that themasses listed are initial mass before burning. . . . . . . . . . . . . . 19
vi
List of Figures
1.1 Firebrand pile forms on a deck during wind-driven firebrand showerexperiments. Figure from Manzello and Suzuki [1]. . . . . . . . . . . 4
2.1 Possible formulation of the energy balance of a firebrand pile on (left)a flat solid fuel, and (right) a solid fuel with a crevice. Based onManzello and Suzuki [1]. . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2 Ignition results of experiments plotted with theoretical ignition bound-ary. Figure from Hadden et al. [2]. . . . . . . . . . . . . . . . . . . . 12
3.1 Side view photograph of radiant heater calibration set-up. The radi-ant heater is on the right side of the figure; gauges are set up on theleft side of the figure. . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.2 Schematic side view of both heat flux experimental set-ups withpropane burner for fabricating firebrands. Figure of burner fromCaton [3]. Not drawn to scale. . . . . . . . . . . . . . . . . . . . . . . 26
3.3 Top view of the experimental board using thin-skin calorimeters withreference numbering of calorimeters (drawn to scale). Thermocouplelocation shown in center of thin-skin array. . . . . . . . . . . . . . . . 27
3.4 Comparison of heat flux calculated using initial calibration and newcalibration after five tests of 12.7 mm firebrands with a 50 g initialpile mass. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.5 Schematic of recipient fuel configurations oriented with respect towind direction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.6 Schematic side view of wind tunnel set-up for thin-skin calorimeterheat flux tests, showing direction of wind and angles of cameras. Notdrawn to scale. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.1 Comparison of heat fluxes measured for a 50 g initial mass 12.7 mmdiameter firebrand pile. . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.2 Energy balance around control volume of a single thin-skin calorimeter. 474.3 Components of heat flux for one radiant heater thin-skin calorimeter
calibration test. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 514.4 Linear fit to correction factor. . . . . . . . . . . . . . . . . . . . . . . 52
vii
4.5 Thin-skin calorimeter temperature as a function of time for 16 thin-skins in array. Thin-skin calorimeter locations can be found in Figure3.3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.6 Total heat flux imparted to the thin-skin calorimeters, calculated us-ing measured thin-skin calorimeter temperatures. . . . . . . . . . . . 55
4.7 Averaged heat flux results for a 12.7 mm diameter firebrand, 9.6 gdeposited mass pile, measured using the water-cooled heat flux gauge. 56
4.8 Magnified replication of previous figure of averaged heat flux resultsfor a 12.7 mm diameter firebrand, 9.6 g deposited mass pile, measuredusing the water-cooled heat flux gauge, showing long-time trends. . . 57
4.9 Plot of averaged heat flux over all repetitions for a 12.7 mm, 9.6 gtest to compare heat fluxes obtained using thin-skin calorimeters andthe water-cooled heat flux gauge. . . . . . . . . . . . . . . . . . . . . 58
5.1 Image sequence showing varying firebrand pile sizes, increasing indiameter from left to right and increasing in pile mass from top tobottom. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
5.2 Temperature as a function of time as measured by thin-skin calorime-ters and thermocouple beneath 12.7 mm diameter firebrand pile of9.6 g deposited mass. . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
5.3 Temperature as a function of time as measured by the thermocoupleand its four immediate nearest thin-skin calorimeters for a 12.7 mmdiameter firebrand pile of 9.6 g deposited mass. . . . . . . . . . . . . 63
5.4 A comparison of temperature as a function of time for the thermocou-ple and an average temperature of the thin-skin calorimeters initiallycovered by the firebrand pile for a 12.7 mm diameter firebrand pileof 9.6 g deposited mass. . . . . . . . . . . . . . . . . . . . . . . . . . 64
5.5 Components of heat transfer for a single thin-skin calorimeter duringa piled test. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
5.6 Components of heat transfer for the covered thin-skin calorimeter ofa single firebrand test. . . . . . . . . . . . . . . . . . . . . . . . . . . 67
5.7 Heat flux as a function of time for a 12.7 mm diameter, 9.6 g de-posited mass test, measured using the thin-skin calorimeter arraywith coverage of thin-skins denoted. . . . . . . . . . . . . . . . . . . . 68
5.8 Averaged temperatures of thin-skin calorimeters based on coverage:full, partial, or no coverage (12.7 mm diameter, 9.6 g deposited masstest). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
5.9 Averaged heat flux of thin-skin calorimeters based on coverage: full,partial, or no coverage (12.7 mm diameter, 9.6 g deposited mass test). 70
5.10 Evolution of spatial distribution of heat flux for a 5 g deposited masspile of 12.7 mm at 100 s, 150 s, 400 s, 900 s, and 1400 s after firebrandsare deposited. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
viii
5.11 Comparison of spatial heat flux maps for (left) a single firebrand, and(right) the largest pile size, 9.6 g deposited mass. From top to bottom,heat flux maps shown an instantaneous heat flux approximately 1 mininto the test, averaged heat flux over the entire test, and maximumheat flux for each location. Due to large heat flux differences, thecolor bar values change for each plot. . . . . . . . . . . . . . . . . . . 72
5.12 A comparison of averaged point measurements from the water-cooledheat flux gauge of 6.35 mm and 9.52 mm diameter firebrands for asingle firebrand. The larger diameter produces a higher heat flux andlasts longer. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
5.13 A comparison of averaged point measurements from the water-cooledheat flux gauge of 6.35 mm and 9.52 mm diameter firebrands for a1 g pile, deposited mass. Differences in heat flux are well within thestandard deviation of the average. . . . . . . . . . . . . . . . . . . . . 75
5.14 A comparison of averaged point measurements from the water-cooledheat flux gauge of 6.35 mm and 12.7 mm diameter firebrands for a2.8 g pile, deposited mass. There is no trend apparent between heatflux and diameter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
5.15 A comparison of averaged point measurements from the water-cooledheat flux gauge of 9.52 mm and 12.7 mm diameter brands for a 5 gpile, deposited mass. The drop-off of the 9.52 mm curve is due toaveraging; the average heat flux curve was calculated only as long asthe shortest test. The larger diameter produces higher heat fluxes. . . 77
5.16 A comparison of averaged heat flux curves obtained using the water-cooled heat flux gauge as deposited pile mass increases for 6.35 mmfirebrand piles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
5.17 A comparison of averaged heat flux curves obtained using the water-cooled heat flux gauge as deposited pile mass increases for 9.52 mmfirebrand piles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.18 A comparison of averaged heat flux curves obtained using the water-cooled heat flux gauge as deposited pile mass increases for 12.7 mmfirebrand piles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.19 A comparison of averaged heat flux curves obtained using the thin-skin calorimeters as deposited pile mass increases for 12.7 mm fire-brand piles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5.20 Peak heat flux as a function of firebrand diameter from thin-skincalorimeter tests. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5.21 Peak heat flux as a function of firebrand diameter from water-cooledheat flux gauge tests. . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
5.22 Peak heat flux as a function of deposited mass from thin-skin calorime-ter tests. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.23 Peak heat flux as a function of deposited mass from water-cooled heatflux gauge tests. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
5.24 Net heating parameter as a function of firebrand diameter for water-cooled heat flux gauge tests. . . . . . . . . . . . . . . . . . . . . . . . 85
ix
5.25 Net heating parameter as a function of deposited pile mass for thin-skin calorimeter tests. . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.26 Net heating parameter as a function of deposited pile mass for water-cooled heat flux gauge tests. . . . . . . . . . . . . . . . . . . . . . . . 87
5.27 Net heating parameter as a function of number of firebrands in thepile. Each marker is an average of all repetitions conducted for aspecific test condition. Error bars are the standard deviation of therepetitions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
6.1 Side view of a flat ignition experiment immediately after firebrand pileis deposited. A small flame occurs in the upper right hand portion ofthe image. Wind direction right to left. . . . . . . . . . . . . . . . . . 90
6.2 Side view of flat ignition experiment over OSB 1 min into experiment.Flames anchored to the fuel surface are visible. The wind directionis from right to left. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
6.3 Side view of a flat ignition experiment over OSB 10 min into theexperiment. Flaming combustion is sustained for the recipient fuel.The wind direction is from right to left. . . . . . . . . . . . . . . . . . 92
6.4 Side view of flat ignition experiment over OSB 15 min into the exper-iment. Flaming combustion has ceased. The wind direction is fromright to left. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
6.5 Conceptualization of possible ignition processes that occur when fire-brands ignite a recipient fuel. . . . . . . . . . . . . . . . . . . . . . . 93
6.6 Visual image (left) and IR image (right) captures for, from top tobottom: 1 min, 10 min, and 15 min into an ignition test. Note thechanged scales on the IR images. The wind is from the right side ofimages. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
6.7 A comparison of the thermocouple temperatures and thin-skin tem-peratures averaged over all covered thin-skins for one test of 10 gdeposited mass. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
6.8 A comparison of the thermocouple temperature with the tempera-tures measured by the four thin-skin calorimeters surrounding thethermocouple for the same test as previous figure. . . . . . . . . . . . 97
6.9 Heat flux as a function of time for the full thin-skin calorimeter arrayfor a 10 g deposited mass test with applied wind. . . . . . . . . . . . 98
6.10 Averaged heat flux curves for 10 g deposited mass tests with wind. . . 996.11 Spatial distribution of heat flux approximately 1 min into a test with
10 g deposited pile mass. Wind direction right to left. . . . . . . . . . 1006.12 Spatial distribution of heat flux over 10 min into the same test with
10 g deposited pile mass. Wind direction right to left. . . . . . . . . . 1016.13 Spatial distribution of heat flux averaged over the first two thirds of
the forced flow test with a 10 g firebrand mass. . . . . . . . . . . . . 1016.14 Spatial distribution of the maximum heat flux obtained for each point
during the forced flow 10 g mass test. . . . . . . . . . . . . . . . . . . 102
x
7.1 A comparison of averaged heat flux curves, measured with thin-skincalorimeters, under ambient and forced flow conditions. The dashedlines represent initial mass 50 g test while the solid lines representinitial mass 100 g tests. All tests use 12.7 mm diameter firebrands. . 106
7.2 Net heating parameter as a function of deposited pile mass for ambi-ent and forced flow experiments using 12.7 mm diameter firebrands. . 107
1 Heat flux as a function of time measured by the thin-skin calorimeterarray under ambient conditions for 6.35 mm diameter firebrands of100 g initial mass. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
2 Heat flux as a function of time measured by the thin-skin calorimeterarray under ambient conditions for 9.52 mm diameter firebrands of100 g initial mass. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
3 Heat flux as a function of time measured by the thin-skin calorimeterarray under ambient conditions for 12.7 mm diameter firebrands of100 g initial mass. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
4 Heat flux as a function of time measured by the thin-skin calorimeterarray under ambient conditions for 12.7 mm diameter firebrands of50 g initial mass. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
5 Heat flux as a function of time measured by the thin-skin calorimeterarray under ambient conditions for 12.7 mm diameter firebrands of20 g initial mass. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
6 Heat flux as a function of time measured by the thin-skin calorimeterarray under ambient conditions for 12.7 mm diameter firebrands of 1brand. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
7 Heat flux as a function of time measured by the thin-skin calorimeterarray under forced flow conditions for 12.7 mm diameter firebrandsof 100 g initial mass. . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
8 Heat flux as a function of time measured by the thin-skin calorimeterarray under forced flow conditions for 12.7 mm diameter firebrandsof 50 g initial mass. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
9 Heat flux as a function of time as measured by the water-cooled heatflux gauge for 6.35 mm firebrands and 6.3 g deposited mass. . . . . . 121
10 Heat flux as a function of time as measured by the water-cooled heatflux gauge for 9.52 mm firebrands and 8.2 g deposited mass. . . . . . 122
11 Heat flux as a function of time as measured by the water-cooled heatflux gauge for 12.7 mm firebrands and a single brand test. . . . . . . 123
12 Heat flux as a function of time as measured by the water-cooled heatflux gauge for 12.7 mm firebrands and 2.7 g deposited mass. . . . . . 124
13 Heat flux as a function of time as measured by the water-cooled heatflux gauge for 12.7 mm firebrands and 5.2 g deposited mass. . . . . . 125
14 Comparison of average heat flux results using the thin-skin calorime-ter array vs. the water-cooled heat flux gauge. 6.35 mm diameter,100 g initial mass. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
xi
15 Comparison of average heat flux results using the thin-skin calorime-ter array vs. the water-cooled heat flux gauge. 9.52 mm diameter,100 g initial mass. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
16 Comparison of average heat flux results using the thin-skin calorime-ter array vs. the water-cooled heat flux gauge. 12.7 mm diameter,100 g initial mass. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
17 Comparison of average heat flux results using the thin-skin calorime-ter array vs. the water-cooled heat flux gauge. 12.7 mm diameter, 50g initial mass. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
18 Comparison of average heat flux results using the thin-skin calorime-ter array vs. the water-cooled heat flux gauge. 12.7 mm diameter, 20g initial mass. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
19 Comparison of average heat flux results using the thin-skin calorime-ter array vs. the water-cooled heat flux gauge. 12.7 mm diameter,single brand. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
xii
Nomenclature
cp Specific Heat (J/kg K)Gr Grashof Number (-)h Convective Heat Transfer Coefficient (W/m2K)k Thermal Conductivity (W/mK)Nu Nusselt Number (-)Pr Prandtl Number (-)q” Heat Flux per Unit Area (W/m2)Ra Rayleigh Number (-)t Time (s)T Temperature (K)
Greek symbolsα Absorptivity (-)δ Thickness (m)ε Emissivity (-)ρ Density (kg/m3)σ Stefan-Boltzmann Constant (W/m2K4)ν Kinematic viscosity (m2/s)
Subscripts0 Initial Value∞ Ambient Valuecond Conductive Termconv Convective Terminc Incident Termindepth In-depth Termmax Maximum Valuenet Net Valueopt Optimumrad Radiative Termrerad Reradiative Termstor Storage Term
AbbreviationsASTM American Society for Testing and MaterialsIR InfraredMC Moisture ContentOSB Oriented Strand BoardNIST National Institute of Standards and TechnologyTSC Thin-Skin CalorimeterWUI Wildland-Urban Interface
xiii
Chapter 1: Introduction
Over the past several decades, the number of devastating wildland-urban in-
terface (WUI) fires has increased drastically. The WUI is defined as the location
where human development abuts or intermixes with wildland vegetation [4]. Fires
that enter these areas are referred to as WUI fires and pose a significant hazard
to homes, people, businesses, and infrastructure. Although the number of wildland
fires has remained relatively constant over the past several decades, the number of
wildland fires affecting human development has increased dramatically. In 2016 over
4,000 structures were destroyed by wildfires in the United States alone. On average,
over 2,000 structures in the United States are lost annually to wildland fires [5].
Global statistics reflect similar trends [6].
A number of factors have influenced the increase in the number of WUI fires:
climate change, fuel management practices, and an increase in human development
into WUI areas [7]. Local changes in climate have occurred in many wildland fire-
prone areas, resulting in hotter temperatures and less precipitation. These changes
affect the likelihood of the initiation and spread of wildland fires. Lower moisture
content (MC) of fuels, as a result of hot temperatures and low precipitation, increases
the severity of wildland fires, as dry fuels are able to ignite more quickly, thus
1
spreading the fire more quickly. In the early 1900s, the United States. Forest
Service introduced a policy to suppress all wildland fires [8]. Although large fires
can result in high structure and economic loss, small fires are a natural part of
the environment and work to mitigate fuel build-up in forests, actually decreasing
the likelihood of large, severe fires. The result of the Forest Service policy was a
huge build-up in fuels across the country. When fires burn in areas with major
fuel build-up, they can often become severe. Finally, by 2000, WUI development
had increased by 50% since 1970, and expansion into WUI areas is expected to
continue [9]. Increased human development in WUI areas increases the number of
structures at risk during WUI fires, as structures are being built in areas that have
traditionally burned without being a threat to human development.
Structures ignite in a wildland fire when the structure is directly or indirectly
exposed to flames or heat from the fire. Three exposure processes are typically
defined for WUI fires: radiation, direct flame contact, and firebrands [7]. Radiant
ignition of structures occurs in a WUI fire when the fire front or another burning
object (e.g. a nearby structure or tree already ignited by the fire) produces sufficient
heat flux over an extended period of time to ignite the structure at a distance.
Although radiant ignition can occur in wildland fires, results of the International
Crown Fire Modeling Experiments found that no ignition of wooden wall segments
occurred when the wall was 20 m or more from an actively crowning 150 kW/m2 fire
[10]. Direct flame contact causes ignition of structures in WUI fires when vegetation
(e.g. ornamental bushes) or other flammable material, such as fallen leaves, forms
an unbroken path from the fire to the structure. Both of these ignition conditions
2
can be mitigated using known methods, such as maintaining adequate defensible
space around a structure [11].
Investigations of past fires, such as the Grass Valley Fire [12], the Waldo
Canyon Fire [13,14], and Witch and Guejito Fires [15,16], have found that firebrands
are responsible for the majority of structure losses in WUI fires [17]. Firebrands, or
burning embers, are small pieces of burning material generated from vegetation or
burning structures during a fire. Firebrands are lofted in the fire plume and can be
transported up to 9 km ahead of the fire front [16]. Firebrands can land in either a
smoldering or a flaming state and cause ignition of a structure several hours after the
main fire front has passed. Even though the evidence shows that firebrands cause
the majority of structure ignitions, firebrands have received less research attention
than radiant ignition has.
Firebrand research that has been conducted can be split into work on three
mechanisms: firebrand generation, lofting and transportation of firebrands, and
deposition and ignition by firebrands [7]. Firebrand generation describes the pro-
duction of embers in a wildland fire and how those embers break off from structures
and vegetation. Firebrand lofting and transport describes the mechanisms which
can loft firebrands from the buoyant plume and transport them over large distances,
reported to be as far as 9 km in some cases [16]. Firebrand deposition and ignition
describes the processes that occur when a firebrand lands on a recipient material.
If this material is combustible, such as a pile of mulch or a deck, ignition describes
the process by which the firebrands ignite the material.
3
1.1 Motivation
The mechanisms governing ignition by firebrands are not well-studied. In
particular, wind-driven firebrand studies by Manzello and Suzuki [1] have found
that firebrands typically form piles and ignition tends to occur in locations with
firebrand piles, as shown in Figure 1.1. Despite the importance of firebrand piles for
ignition, most previous small-scale work has focused on ignition by a single firebrand.
A number of large-scale studies (full-scale structures or structural components) exist
which investigate ignition propensity from piles [37]; however, few small-scale studies
using firebrand piles can be found in the literature (e.g. [18]).
Figure 1.1: Firebrand pile forms on a deck during wind-driven firebrand shower
experiments. Figure from Manzello and Suzuki [1].
During a wildland fire, firebrands can ignite vegetative fuels, such as leaves or
grasses, or WUI fuels, such as decks, porches, fences, or other dense fuels. Previous
studies on structure loss have found that ignition by firebrands is often indirect (i.e.
4
firebrands ignite an adjacent structure, such as the WUI fuels described, and that
adjacent structure subsequently ignites the structure). Previous work on firebrand
ignition has focused primarily on vegetative fuels, which are more porous than solid
WUI fuels. The porosity of vegetative fuels allows firebrands to embed themselves in
the recipient fuel, creating greater contact between the ignition source and fuel. On
the other hand, the contact area between solid fuels and firebrands is much smaller.
Firebrands accumulate in piles on top of solid fuels. The behavior of firebrand piles
is expected to be different than that of a single firebrand because the heating from
individual brands in the pile can interact. The pile behaviour, as well as the fact that
the fuel is solid rather than porous, is likely to change the heat transfer mechanisms
governing the smoldering ignition process and the transition from smoldering to
flaming ignition.
This work aims to fill some of the gaps in firebrand ignition research. One part
of the research is to study the heat fluxes produced by piles of smoldering firebrands
in order to understand the contact heat transfer between firebrands and the recipient
fuel. No data currently exists quantitatively connecting pile sizes (measured by the
mass or number of firebrands) and ignition; however, qualitative results of other
studies [18] suggest that greater numbers of firebrands are significantly more likely
to ignite a recipient fuel than are single firebrands. The work aims to determine
whether there is a link between pile size and ignition and describe the process of
ignition of typical WUI fuels.
5
1.2 Objectives
The ultimate goal of this research is to understand the ignition process of dense
structural WUI fuels by firebrand piles. Ultimately, understanding the ignition
process of these fuels can allow for the development of a model of the ignition
process. There are three main objectives in this work.
In order to measure the heating from and thermal characteristics of firebrand
piles, it is necessary to have a method to reliably conduct pile experiments and
measure relevant thermal quantities. The first objective of this study was thus
to formulate a methodology for measuring temperatures and heat fluxes in exper-
iments using smoldering firebrands. This objective necessitated the development
of a reproducible and repeatable method for the creation of firebrand piles and a
reliable method of measuring heat fluxes from firebrand piles to different substrates,
including identifying the best sensor to measure the heat fluxes in firebrand piles.
The second objective of this study was to quantify the heat flux from firebrand
piles to an inert surface, in order to isolate the heat flux produced by firebrands from
the heat flux produced when a recipient fuel ignites. This objective was feasible after
the first objective had been completed. This objective included determining how
parameters such as firebrand diameter and pile size (mass, number) affect changes
in heating behavior.
The final objective was to describe the ignition process and the thermal char-
acteristics (heat flux, temperature) at the time of flaming ignition. The objective
was meant to connect ignition and heat fluxes measured for the second objective.
6
Conducting firebrand pile experiments on a recipient fuel and on an inert substrate
allowed for a description of ignition conditions.
7
Chapter 2: Literature Review
2.1 Overview
The main goal of this project is to understand the ignition process by which
firebrand piles ignite recipient WUI fuels. In this chapter, an overview of ignition
is provided in order to form a basis for understanding the firebrand pile. The
differences between radiant ignition and ignition via firebrand piles are discussed, as
are the differences between flaming and smoldering ignition. The chapter concludes
with a discussion of previous experimental studies of firebrand ignition.
Ignition is the initiation of either flaming or smoldering combustion. This
process is associated with a temperature rise of a fuel to a certain critical point. In
the case of flaming ignition, the temperature rise results in pyrolysis of the solid fuel
to gaseous fuel. At a critical mass flux, this gaseous fuel will ignite. In the case of
smoldering ignition, combustion takes place in the solid phase. An important aspect
of this process is how energy, in the form of heat, is transferred to a fuel in order to
initiate a temperature rise.
The heat transfer process for a firebrand pile is not completely understood;
however, Figure 2.1 shows hypothetical formulations of the possible energy balance
of a firebrand or firebrand pile on a recipient fuel. Firebrands ignite recipient fuels
8
due to heat transfer across an interface. Depending on how well the firebrand
contacts the recipient fuel, which may be poor for solid fuel but relatively good for
porous fuels, conduction or radiation will dominate the heat transfer process. This
heat transfer will occur in a variety of modes that ultimately result in in-depth heat
transfer to the recipient fuel, q′′indepth, eventually heating the fuel until it ignites. The
pile will also produce heat due to continued smoldering, proportional to the rate of
mass loss and latent heat of the fuel, m′′L; however, the pile will also lose heat due
to convection, q′′conv and radiation q′′rad to the ambient. For situations when there is
a gap in the fuel, it is suspected there could be additional re-radiation within the
fuel bed and access to oxygen that promotes surface oxidation, making smoldering
ignition and transition to flaming easier.
Figure 2.1: Possible formulation of the energy balance of a firebrand pile on (left)
a flat solid fuel, and (right) a solid fuel with a crevice. Based on Manzello and
Suzuki [1].
2.2 Ignition: Flaming and Smoldering
Descriptions of flaming and smoldering ignition are provided. Ignition theories
applicable to the current problem are reviewed. A description of flaming ignition and
9
ignition by radiation is described because these areas have received more research
attention. Understanding the ways in which researchers formulate and describe
radiant ignition can provide context when formulating a way to describe ignition
by firebrand piles. A brief review of smoldering combustion and its thermal char-
acteristics is provided in order to give a comparison to the magnitude of thermal
characteristics measured in this study’s heat flux and ignition experiments.
There are several methods of describing a critical ignition point for radiant
ignition. Radiant ignition can occur as a result of a sustained ignition temperature
[19], sustained heat flux over a given period of time [10], or a certain mass flux over
a given time [20].
While ignition thresholds such as those described above are typically used to
determine flaming auto-ignition, the initiation of smoldering combustion is partic-
ularly important in our application. Smoldering is a solid phase combustion pro-
cess, a heterogeneous oxidation reaction, which peaks at lower temperatures and
heat release rates than flaming combustion – approximately 450–700°C and 10–30
kW/m2 [21].
In addition to producing lower peak temperatures, critical thresholds for smol-
dering combustion are also lower. For example, Rein found that under an incident
radiant heat flux, smoldering ignition of polyurethane could be initiated at 7 kW/m2,
while spontaneous flaming ignition did not occur until 30 kW/m2 [21]. Anthenien
and Fernandez-Pello were able to ignite polyurethane using conduction at as low
as 3.1 kW/m2 [22]. Although these values are for polyurethane, which has differ-
ent thermal properties than WUI fuel materials (woods, plastics, and wood-plastic
10
composites), the comparison between values for critical heat flux for smoldering
and flaming ignition are notable as the values for smoldering are significantly lower.
Gratkowski et al. conducted experiments on the smoldering ignition of plywood in
a cone calorimeter. They found that the minimum heat flux at which the plywood
would ignite was 7.5 kW/m2 [23].
2.2.1 Theories of Ignition
A number of theories of ignition have been developed which are related to the
problem of firebrand ignition, although some of these theories have been developed
for other related applications. Gol’shleger et al. provide a theory of hot spot ignition,
which assumes a hot, inert object lands on a reactive material. The object is assumed
to create pockets of heating in the recipient fuel, and these hot areas are assumed
to be the locations where ignition occurs [24]. Although this theory could be a
good starting point, the assumption of an inert object is not valid for the firebrand
problem.
Hadden et al. [2] performed experiments using hot metal particles to ignite cel-
lulose fuel beds. They varied the particle size and found a hyperbolic relationship
between the particle size and particle temperature required to ignite the fuel bed.
They applied hot spot theory to their data and obtained a qualitative fit. Results
of the experimental study and the theoretical fit can be seen in Figure 2.2. Further
experiments by Zak et al. [25] and Fernandez-Pello et al. [26] have noted the impor-
tance of heat losses from larger particles. The hot spot theory, though qualitatively
11
validated, does not take into account ongoing reactions in the hot particles, which
may play an important role in firebrand pile heating.
Figure 2.2: Ignition results of experiments plotted with theoretical ignition bound-
ary. Figure from Hadden et al. [2].
Gray also discusses the concept of criticality, hypothesizing that critical igni-
tion curves can by plotted where the heat loss from a hot object balances with the
heat generation within the fuel [27]. This concept is similar to the ignition boundary
plotted by Hadden et al. [2]. This concept could be important in terms of defining
ignition criteria for firebrand ignition of solid fuels.
In a different analysis, Gratkowski et al. employed the Frank-Kamenetskii
parameter, a measure of thermal stability, when using self-heating ignition theory.
This parameter indicates the balance of heat generation in relation to heat losses
and whether the heat generation will be sufficiently larger to cause thermal runaway
of the system [23]. The theory found good agreement with experiments of radiant
smoldering ignition of plywood and with self-heating theory.
12
Another theory of ignition is to solve the energy equations of a smoldering re-
action in order to determine a critical ignition heat flux as a function of smoldering
depth. This method is described in Dosanjh and Pagni [28] and applied experi-
mentally to the smoldering ignition of polyurethane in Anthenien and Fernandez-
Pello [22]. Using this method, Anthenien and Fernandez-Pello found good agreement
between the experiments and the theoretical prediction of ignition as a function of
critical heat flux and time.
2.3 Experimental Studies
A number of previous studies have explored facets of the firebrand ignition
problem, both experimentally and theoretically. A review of these studies provides
background on current experimental methods which can be a useful comparison for
formulating the research methods used for the current study.
The following sections describe experimental studies on ignition by firebrands.
Specifically, different firebrand production methods are discussed, and ignition ex-
periments on vegetative and solid WUI fuels are reviewed to highlight important
considerations when formulating the experiments in this study.
2.3.1 Firebrand Production
A review of firebrand production methods used in previous studies was con-
ducted before determining the method used in the current study. Below are concise
descriptions of several firebrand production methods. Ellis [?] fabricated glowing
13
firebrands from eucalyptus bark by igniting the sample using a gas stove and then
placing it in a “truncated cone”, in which the sample rotated while exposed to
airflow. This method was used only for a single firebrand. Santamaria et al. [30]
simulated firebrands both with charcoal and by submerging virgin material into
flames from a heptane pool fire. In a method which would later be used in the Na-
tional Institute of Standards and Technology (NIST) firebrand generator, called the
NIST Dragon, Manzello et al. suspended small Ponderosa Pine (Pinus ponderosa)
disks above butane flames to produce either flaming or glowing firebrands [18]. Af-
ter a set time, the firebrands were exposed to wind and became part of a firebrand
shower.
Dowling [31] ignited wooden cribs and set them adjacent to his experiment. A
similar procedure was used by Quarles et al. [32] when they tested decks for firebrand
exposure by igniting a wood crib on top of the deck. Other methods for producing
flaming firebrands include igniting virgin fuel (e.g. wood, bark) with a flame and
allowing it to burn [33]. None of the studies reported time in flames or whether
the time was held constant for all tests; however, the use of flames for firebrand
production is a common and successful option. A refinement on this method was
utilized in this study.
2.3.2 Firebrand Ignition of Vegetative Fuels
A collection of experimental studies exist which focus on ignition of fuels by
firebrands. A number of these studies focus on recording Ignition vs. No Ignition
14
for a range of experimental parameters; however, these studies do not investigate
heat fluxes associated with ignition or the mechanisms governing the transition
from smoldering to flaming ignition. A large number of studies on firebrand ignition
utilize vegetative fuels as the recipient fuel. Many studies also use one firebrand or a
small number of firebrands, rather than large piles. Of these studies, many explore
ignition as a function of number of firebrands deposited. There are a number of
larger scale studies using solid fuels as the recipient fuel for ignition experiments.
Manzello et al. [18] used either a single brand or three to four firebrands to
test ignition of pine needles, shredded paper, and cedar wood crevices. A single
glowing firebrand was able to initiate smoldering in the completely dry shredded
paper under an external airflow of 0.5 - 1.0 m/s. Four large (50 mm diameter disk)
firebrands were able to ignite pine needles, which transitioned from smoldering to
flaming under a 1.0 m/s wind. No glowing firebrands were able to ignite the cedar
crevice. In later tests, Manzello et al. [34] conducted similar tests as described above
on three other fuels: shredded hardwood mulch, pine straw mulch, and cut grass.
A single glowing firebrand was never able to ignite any of the fuels. Four glowing
50 mm disk firebrands were able to initiate smoldering ignition for some of the fuel
beds under external airflow.
Viegas et al. [33] conducted ignition tests using bark and pine cones as fire-
brands and six fuel beds of dried and live eucalyptus leaves, dried and live pine
needles, hay, and straw. Experiments were conducted under ambient conditions.
For a single glowing firebrand, no ignition occurred. No experiments of glowing
firebrands under external airflow conditions were conducted.
15
Ellis [29] conducted experiments using glowing and flaming firebrands made
of 500 mm long bamboo pieces. The moisture content of the fuel bed, eucalyptus
litter, was varied from 4-21% and external airflow varying from 0 - 2 m/s was applied.
Glowing firebrands only produced ignition for fuel beds with lower moisture contents.
In general, fuel beds with higher moisture content required higher wind to produce
ignition.
2.3.3 Firebrand Ignition of Solid Fuels
Early tests by Waterman and Takata [35] were the first experiments using
firebrands on urban fuels. They found that ignition probability increased with in-
creasing wind speed. An additional external heat flux was applied to act as a pilot
ignition source. Dowling [31] conducted firebrand tests in bridge timbers and found
that 7 g of firebrands were sufficient to produce ignition.
A series of experiments by Manzello et al. have been conducted on both a small
and large scale using the NIST Dragon. In small-scale tests, they deposited glowing
firebrands (between one and four total firebrands) in a cedar crevice. External
airflows of 0.5 m/s and 1.0 m/s were applied. No ignition occurred [18]. In further
experiments, four glowing firebrands were deposited into crevices made of either
plywood or oriented strand board (OSB), both common building materials [36].
Ignition was sensitive to angle – only tests at 60°or 90°ignited. Additionally, of
tests at these angles, only tests with 2.4 m/s external airflow and dry recipient fuel
16
ignited. The authors expected that the fuels with higher moisture content (i.e. 11%)
did not ignite due to the higher thermal inertia of these samples [36].
Manzello et al. [37] provides an extensive summary of large scale experiments
in Japan’s Building Research Institute Fire Research Wind Tunnel Facility using
the full-scale NIST Dragon. These experiments determined vulnerabilities of siding,
roofing, and other portions of full structures that are susceptible to firebrand attack.
Santamaria et al. [30] conducted ignition tests on solid wood boards using a
heat flux from a controlled source heater and piles of charcoal to simulate firebrands;
however, charcoal was not found to simulate firebrands well. Heat fluxes from bark
firebrands were taken on a vermiculate inert board and found to produce heat fluxes
of up to 6 W/m2.
17
Chapter 3: Experimental Methodology
3.1 Overview
Several experiments were conducted to determine characteristics of piled fire-
brand behavior. There were three main objectives of this project, which produced
three overall sets of experiments:
1. Calibration tests used to identify and quantify given measurement techniques
2. Heat flux tests with firebrand piles under ambient conditions
3. Tests under forced flow conditions to describe an ignition condition
Each set of experiments can be broken down into individual test configurations
based on the measurement instruments used. This further breakdown of experimen-
tal sets will be described in each section. While the calibration tests were focused on
the types of instruments to be used and correct quantification of heat flux, the other
two sets of experiments measured quantities (i.e. temperature, heat flux, ignition)
in relation to firebrand pile sizes. Table 3.1 shows the test matrix of the firebrand
pile sizes in terms of diameter and mass used for each specific experiment.
18
Table 3.1: The test matrix. For each type of test, experiments were conducted using
the diameter and masses of firebrands indicated. Note that the masses listed are
initial mass before burning.
Diameter Ambient
Water-Cooled
HFG
Experiments
Ambient TS
Calorimeter
Experiments
Forced Flow
TS
Calorimeter
Experiments
Ignition
Experiments
6.35 mm 1 brand, 20 g,
50 g, 100 g
100 g None None
9.52 mm 1 brand, 20 g,
50 g, 100 g
100 g None None
12.7 mm 1 brand, 20 g,
50 g, 100 g
1 brand, 20 g,
50 g, 100 g
50 g, 100 g 50 g, 100 g
19
3.2 Calibration Tests
The specific measurement techniques chosen, as well as calibration results,
are described in Chapter 4. The first objective of identifying a reliable method
of measuring thermal characteristics necessitated the identification of a method of
measuring heat flux from firebrand piles. An accurate instrument for measuring heat
flux was difficult to find, as most heat flux measurements are focused on radiative
fluxes. Three potential heat flux gauges were identified as options for our heat flux
experiments: water-cooled heat flux gauges, thermopiles, and thin-skin calorimeters.
For a full description of each instrument used, please see Chapter 4.
The water-cooled heat flux gauge is the most commonly used method for
measuring heat flux in fire protection engineering research. The main concern with
the use of a water-cooled heat flux gauge for this application was whether the water-
cooling for the gauge would cool the firebrands. Testing whether cooling occurred
was one subset of the calibration tests.
Thermopiles have the advantage of not requiring water-cooling; however, they
can become fragile at high temperatures (including those temperatures expected
from firebrand piles). Due to the stochastic nature of ignition, it was desirable to
have an instrument sufficiently durable to withstand many repetitions of heat flux
and ignition experiments. For this reason, a thermopile was not employed.
The thin-skin calorimeter is more durable than the thermopile and also does
not require cooling, making it a potentially good option for firebrand pile experi-
ments where heat fluxes are low enough that cooling could likely influence experi-
20
ments. Unlike water-cooled heat flux gauges and thermopiles, thin-skin calorimeters
measure temperature which must be converted to heat flux via heat transfer calcula-
tions. An understanding of the energy balance for the complex heat transfer around
the firebrand pile posed the largest challenge for using the thin-skin calorimeters for
this application. Calibration of thin-skin calorimeters was the second subset of the
calibration tests.
A thin-skin calorimeter is a calorimeter fabricated by welding a thermocou-
ple to the backside of a thin metallic sheet. Given the properties (density, specific
heat, thickness) of the metal, the thin-skin calorimeter can be used to measure heat
transfer rates by calculating heat transfer components from a measured tempera-
ture history, using the assumptions of one-dimensional heat transfer and lumped
capacitance. The test method is described in ASTM E459 [46].
ASTM E459 states one application of thin-skin calorimeters as “heat transfer
measurements in fires and fire safety testing” [46]. A study by Hildalgo et al. [38]
describes an application of thin-skin calorimeters for measuring incident radiant
heat flux in large-scale fire tests. This study describes a method of calibration of
thin-skin calorimeters meant to better account for uncertainties in the heat transfer
calculations.
In order to determine whether the water-cooled heat flux gauge cooled fire-
brands, two sets of tests were conducted: one set with a 2.54 cm (1.0 in) Medtherm
heat flux gauge and the second set with a smaller, 1.27 cm (1/2 in) Medtherm heat
flux gauge. For both gauges, the measurement area is the same size; however, the
cooling area is larger for the larger gauge. Both gauges were used for repetitions of
21
firebrand pile tests using a single brand and 50 g initial mass for three diameters of
firebrands. Averages for each diameter and pile mass were calculated and the two
gauges compared. Results can be found in Chapter 4, Section 4.2.1.
Initially, several methods were employed to attempt calibration of the thin-skin
calorimeters. One method of conductive calibration was attempted. This method
consisted of heating a copper slug in a propane flame until it was glowing and
then placing the slug on the thin-skin calorimeter. This method was not continued
because of uncertainty surrounding the contact resistance between the slug and
the thin-skin calorimeter. Another method attempted was calibration using a cone
calorimeter. Due to availability constraints on the cone calorimeter, a radiant heater
was used for calibration following the method described below.
Ultimately, thin-skin calorimeters were calibrated following a method described
by Hildalgo et al. [38]. The calibration used a radiant propane heater, a Dyna-Glo
TT15CDGP 15,000, which attached onto the top of a propane tank. The heater
was oriented vertically and placed directly facing a reference water-cooled heat flux
gauge and the thin-skin calorimeter to be calibrated. In order to attain higher heat
fluxes, the steel heat reflector was removed from the front of the heater. The ref-
erence gauge and the thin-skin calorimeter to be calibrated were set up using ring
stands and were on a level vertically and horizontally. Additionally, it was critical
that the surface of each gauge was the same distance from the heater. The gauges
were set so that the center point between the gauges was in-line with the center of
the radiant heater. Both the heater and the gauges were beneath a flow hood in
order to exhaust products.
22
During calibration, the gauges were set as far back as possible to one side
of the hood, while the radiant heater was set to the other side of the hood (see
Figure 3.1. The radiant heater was kept on “High” for all calibration points and
was moved closer to the gauges in order to increase the heat flux. Each position
was kept constant for 10 min in order to account for the time lag of the thin-skin
calorimeter and allow several minutes of constant readings. At the end of 10 min, the
radiant heater was moved closer to the gauges by a few centimeters. This procedure
was repeated until the radiant heater surface was as close as possible to the gauges
without the gauges entering the thermal boundary layer of the heater. A total of
five to fifteen radiative heat fluxes were used for each calibration.
Figure 3.1: Side view photograph of radiant heater calibration set-up. The radiant
heater is on the right side of the figure; gauges are set up on the left side of the
figure.
23
Several thin-skin calorimeters were calibrated by this method in order to de-
termine what differences, if any, there were between specific thin-skin calorimeters.
A total of five thin-skin calorimeters with similar emissivity were calibrated and
compared to determine a general calibration.
3.3 Heat Flux Tests With Firebrand Piles
The second overall set of experiments was to quantitatively determine the heat
fluxes from firebrand piles and to test the calibration methods employed as part of
the first set of experiments. For this set of experiments, two subsets of analogous
tests (see test matrix in Table 3.1) were conducted: one using a water-cooled heat
flux gauge and another using an array of thin-skin calorimeters. Using both methods
of heat flux measurement allowed for a comparison of the two methods and a check
for the thin-skin calorimeter calibration. These tests were all conducted on an inert
ceramic insulation board in order to isolate the heat flux from the firebrand piles.
Experiments using thin-skin calorimeters provided spatial maps of the heat
flux beneath the firebrand pile and indicated the total area heated by the firebrands,
as well as heat fluxes later in the tests. Experiments with a single-point water-
cooled heat flux gauge provided time-resolved heat flux measurement early in each
experiment and were used to validate the thin-skin calorimeter calibration.
24
3.3.1 Experimental Design
Two experimental set-ups were built for heat flux tests, and differences in
design arose out of the different size and shape requirements of the different sensors.
Both set-ups, shown in Figure 3.2, were created by layering a 1.5 cm thick sheet of
Superwool 607 High Temperature ceramic insulation board atop a 1.2 cm thick piece
of plywood. The insulation board was used as an inert material and the plywood
was used to provide a stabilizing mass for the set-up under wind. The insulation and
plywood, which comprised the whole experimental board, were both 18 cm x 18 cm.
Due to the sensitivity of the load cell and the ambient air movement of approximately
0.1 m/s, any overhang of the experimental board increased uncertainty in the mass
measurements; thus, the size was chosen to match the size of the load cell test
surface. Although no mass readings were taken during the water-cooled heat flux
experiments, the set-up size was kept constant in order to decrease potential changes
in airflow around the board.
In the thin-skin calorimeter set-up, sixteen sensors were arranged in a 4 x 4
array (see Figure 3.3) with the centers of the calorimeters separated by 1.5 cm,
leaving 0.5 cm between the edges of adjacent thin-skins. The wires from the sixteen
thin-skin calorimeters were threaded through the insulation and through drilled
holes in the plywood, then secured onto the back side of the plywood. In this set-
up, a Chromel-Alumel K-Type thermocouple with 30 gauge wire was placed on top
of the insulation and secured using a small metal staple, so that the thermocouple
bead was directly atop the insulation and in the middle of the array of thin-skins.
25
Figure 3.2: Schematic side view of both heat flux experimental set-ups with propane
burner for fabricating firebrands. Figure of burner from Caton [3]. Not drawn to
scale.
Thermocouples were fabricated in the laboratory using a HOT SPOT TC Welder
spot welder. The thermocouple was placed on top of the insulation, beneath the
firebrand pile. It measured the temperature of either the gaseous products within
the firebrand pile or the temperature of a firebrand, if a particular firebrand landed
on the surface of the thermocouple. Thus, the temperature measured is a ballpark
pile temperature, but may experience variation based on how the firebrands in the
pile land.
The entire set-up was placed on a load cell and inside a laminar flow hood
for ambient experiments. The airflow in the laminar flow hood above the surface
of the experimental board was measured to be 0.1 m/s using a hand held hot wire
anemometer placed at several locations around the experimental sensors.
26
Figure 3.3: Top view of the experimental board using thin-skin calorimeters with
reference numbering of calorimeters (drawn to scale). Thermocouple location shown
in center of thin-skin array.
During thin-skin calorimeter heat flux experiments, a Mettler Toledo MS6002TS
load cell with ± 0.01 g precision was used to measure the mass loss of the experi-
ment. Despite this precision, ambient airflow caused the mass readings to fluctuate
by approximately 1 g. Due to the low mass loss of the firebrand piles (on the order
of 1 - 8 g) and the airflow fluctuations, reliable mass loss rates were not calculated.
For the water-cooled heat flux gauge experiments, the 1.27 cm Medtherm
water-cooled heat flux gauge was used for this set of experiments. In the water-
27
cooled heat flux gauge set-up, a hole 1.27 cm in diameter, the size of the casing of
the water-cooled heat flux gauge, was drilled into the wood and insulation board.
The water-cooled heat flux gauge was placed directly in the center of the insulation
board. The tubes from the gauge were L-shaped and were secured to the top of the
plywood beneath the insulation for stability of the gauge. Water tubes for cooling
were connected to a pump in a 15 gal water storage tank. A thermocouple was placed
in the water tank in order to ensure a consistent cooling source was provided. In
both this set-up and the previous set-up, a video camera recorded the experiment
from an elevated side view.
3.3.2 Wood Properties and Firebrand Production
Birch wooden dowels were used to produce firebrands used in all tests. For
these experiments, cylindrical dowels of three diameters, 6.35 mm, 9.52 mm, and
12.7 mm, were cut to 25.4 mm in length. While firebrands come in different shapes
and sizes – disks or bark pieces, needles, and sticks – collection studies by Manzello
et al. [39] found that cylindrical sticks are one of the most common brand shapes.
Although Santamaria et al. [30] also found bark to be a common firebrand, sticks
are more reproduciblely simulated. Diameters and length were chosen because they
were near the average size based on collection studies and because of the availability
of dowel sizes. Although 12.7 mm diameter firebrands may be larger than average,
this provided the ability to compare over a range of diameters (from 6.35 mm to 12.7
mm). It is important to note that these diameters represent the initial diameters of
28
the wood; however, the diameters of the burned firebrands are smaller. See Caton [3]
for representative proportions of diameter decrease in wooden dowels.
All dowels were birch, which was chosen in this study as it was readily avail-
able in the sizes of firebrands typically observed in WUI fires; however, its den-
sity is slightly higher than that of softwoods more commonly found in WUI fires
(approximately 600 kg/m3 in comparison to the 350-530 kg/m3 density range of
softwoods [40]). Wood was dried at 103 ± 2°C in a VWR Gravity Convection
Oven per ASTM Standard D4442 [41]. Drying was complete when the mass of the
sample steadied within ± 0.1 g for three consecutive hours, when measured on a
Mettler Toledo NewClassic MF load cell (Model # MS32001L) with accuracy ±
0.1 g. Moisture content (MC) was also measured using an A&D MF-50 moisture
content analyzer. Dried wood was placed in one gallon plastic bags with drierite to
keep it from regaining moisture between drying and testing. The wet-basis MC of
the samples was in the range of 0 ± 0.5 %.
In order to test the effect of smoldering firebrands on recipient fuels, it was
necessary to devise a method to repeatably produce smoldering firebrands. Previous
studies using firebrands have used a variety of methods of producing brands, but
many focus on flaming rather than smoldering brands. Of those studies that used
glowing firebrands, methods used to produce smoldering firebrands were either not
repeatable or difficult for use in producing a pile of firebrands. Previous firebrand
production methods are discussed in Section 2.3.1.
Several fabrication methods were tested to determine the most repeatable
method of producing smoldering firebrands. Initially, wooden dowels were placed in
29
a 12 cm x 16.5 cm x 11 cm Thermolyne furnace (Type 1400) at a set temperature
between 300-450°C for a given time between 5-30 min. (The temperatures and times
were varied to determine if a given temperature-time combination would produce
a pile of glowing firebrands.) Although one to two dowels typically glowed at their
tips, most dowels were either charred or had turned to ash when removed from the
oven. It was determined that the oven did not provide sufficient oxygen for the
dowels to reach a smoldering state [21].
Another method tested for producing firebrands was heating on a Corning PC-
600D Hotplate. Dowels were deposited on the hot plate and turned occasionally.
Imaging using a Seek Thermal IR camera determined that the hot plate did not
provide even heating to the brand surface. Additionally, the highest temperature
produced by the apparatus (550°C) did not provide sufficient heat for the wood
to smolder. It was determined that exposure to flames would likely be necessary
to produce smoldering and would more realistically replicate exposure during a
wildland fire.
In order to best simulate a real fire exposure, dowel pieces were exposed to
flames. Initially, dowels were put in a small metal bowl with 5 mL of heptane or
ethanol; however, though the dowels flamed, they did not continue glowing after
the flames ceased. Ultimately, the dowels were exposed to a propane flame. All
dowel pieces to be used in a given test were put in a wire mesh basket which was set
on a Bunsen-type burner over a propane flame (e.g. for a test of 25 g of 6.35 mm
brands, all 25 g would be exposed to flame simultaneously). The flame height was
kept constant, so that it touched the bottom of the basket in every test. The flame
30
was turned off after all of the brands had visibly ignited, approximately 150-200 s.
After the cessation of flaming, glowing firebrands were deposited on the test set-up.
Another method of extinguishing was tested, in which the flames were manually
extinguished by placing the firebrands in a metal box with a hole drilled in the top
to allow smoke to escape. It was found that brands were more reproducibly formed
in a glowing state when they self-extinguished. This firebrand production procedure
was kept constant throughout all of the tests.
Although this firebrand production method was repeatable to conduct, there
were certain associated uncertainties. Larger quantities of firebrands flamed for
longer periods of time after the burner was turned off. As a result, larger piles of
firebrands may have been in a more degraded state upon deposition than smaller
piles or single firebrands. It is uncertain whether these differences produced any
discernible differences in the data recorded.
3.4 Experimental Procedures
A set of experimental procedures was developed for each type of test, depend-
ing on the instruments used and the flow conditions (forced flow or ambient). All
general procedures were kept the same for all types of tests. After the experimental
board was in place, firebrand production was begun. Once the burner was turned
off, data acquisition was initiated in order to capture ambient conditions before the
test began. Data acquisition used in each test is listed in Table ??. Glowing fire-
brands were deposited on the test board, and data acquisition was continued until
31
measured parameters decreased below a set threshold, specified for each instrument
used. For tests in ambient conditions, both the propane burner and the test board
needed to fit under the laminar flow hood, while forced flow and ignition experi-
ments were conducted in a large burn room with overhead exhaust capabilities. Due
to the proximity of the propane burner to the test set-up in the ambient configu-
ration, the flaming of the firebrands elevated baseline temperatures of the thin-skin
calorimeters and thermocouples to approximately 30°C before the tests began.
3.4.1 Firebrand Pile Tests Using Thin-Skin Calorimeters
Before beginning each test using the thin-skin calorimeter array, each calorime-
ter was tested to determine that it had not broken. This procedure was undertaken
because the thin-skin calorimeters were fragile – sometimes the attachment of the
wire to the metal broke. Testing was conducted by exposing each calorimeter to a
small butane flame and checking whether it increased in temperature as expected.
To begin the experiment, firebrands were placed onto the thin-skin calorimeter
array after they ceased flaming. For tests with single firebrands, tongs were used to
move the firebrand from the wire mesh basket and to set it directly atop Thin-Skin
Calorimeter 6 (see Figure 3.3). For piles, glowing firebrands were poured onto the
center of the test set-up from the side of the wire mesh basket, using a large pair of
tongs as a guide. Once the firebrands were deposited on the test set-up, tongs were
used to push the firebrands to the center of the board if they had fallen to the side.
For larger piles of 6.35 mm firebrands, a few glowing brands sometimes decreased
32
in diameter sufficiently to fall through the holes in the mesh. When this occurred,
those firebrands were not later placed on the pile.
After firebrands were deposited on the test board, the pile was left, and data
was recorded until the thermocouple and thin-skin calorimeters decreased below
27°C. This cut-off temperature for thermocouples was used because heat fluxes cal-
culated at this temperature were small. At this point, data acquisition was stopped.
The firebrand remnants, char and ash, were swept off the test board using a small
piece of cardboard, and the sensors were cleaned of debris using compressed air.
3.4.2 Firebrand Pile Tests Using Water-Cooled Heat Flux Gauge
For tests using a water-cooled heat flux gauge, water tubes were connected to
the gauge and the pump was placed in the 15 gal water storage container and allowed
to cycle water for two to three minutes before firebrand production began. For tests
using a single firebrand, the brand was placed directly atop the center of the sensor.
For piles, the same procedure was followed as for the thin-skin calorimeter array.
Data acquisition was stopped when the heat flux gauge no longer registered any
external heat flux.
After each test, the gauge was checked to see if any paint had been discolored or
scratched. If significant changes occurred, the heat flux gauge was repainted and re-
calibrated, as described in Section 4.2. In order to ensure that the firebrands on top
of the sensor did not damage the gauge, re-calibration was conducted after the first
five heat flux tests were finished. Both the old and new calibration coefficients were
33
applied to voltage measurements on the final test, so that the effect of calibration
could be determined. The results of this comparison are shown in Figure 3.4. The
differences in the heat fluxes produced by the different calibration coefficients were
within the uncertainty of the tests. The procedure was repeated again after fifteen
more tests. Again, the effect on heat flux was sufficiently small.
Figure 3.4: Comparison of heat flux calculated using initial calibration and new
calibration after five tests of 12.7 mm firebrands with a 50 g initial pile mass.
34
3.5 Tests Under Forced Flow Conditions
The objective of conducting tests under forced flow was to describe a spe-
cific ignition condition. The application of wind was necessary to aid ignition, as
preliminary tests of flaming ignition of a recipient fuel under ambient airflow con-
ditions were not successful. Again, two subsets of tests were conducted. The first
subset was a series of ignition tests using firebrand piles on a recipient fuel which
were conducted to determine time to ignition and the necessary wind velocity and
pile size to produce flaming ignition. The second subset was a series of heat flux
measurements on an inert board using the same external conditions as were used to
produce ignition. This second subset of experiments was conducted to isolate the
thermal characteristics that occurred immediately preceding ignition.
3.5.1 Ignition Test Set-up
The experimental set-up for the ignition tests was fairly straightforward. A
6.35 mm thick aluminum plate with a super elliptical leading edge (described in
[42]) was placed several centimeters in front of the outlet of the wind tunnel. The
aluminum plate allowed for the formation of a laminar boundary layer before the
firebrand test section. In the ignition tests, an 18 cm x 18 cm sheet of oriented strand
board (OSB) was placed behind the aluminum edge, flush with its surface. (The
analogous thin-skin calorimeter set-up shown in Figure 3.6 shows the orientation of
the aluminum plate in relation to the wind and experimental set-up.)
35
OSB was used in these tests because of its common usage as a building mate-
rial, often used as sheathing for walls and roofing. It has also been used for ignition
tests in other studies [36]. OSB was dried at 103 ± 2°C in a VWR Gravity Convec-
tion Oven, following the same procedure as was used to dry the wood for firebrands.
As described previously, dried wood was placed in gallon plastic bags with drierite
to keep it from regaining moisture between drying and testing. Drying the wood was
important, as many studies have found that the ignition of recipient fuels (including
woods and plastics) is sensitive to the MC of the fuel [18,33,36,43].
Preliminary ignition tests were conducted under ambient conditions to de-
termine whether flaming ignition could be achieved. Firebrand pile sizes up to
100 g initial mass (10 g deposited mass) of 12.7 mm firebrands were tested; how-
ever, no piles achieved flaming ignition, though most achieved smoldering ignition.
The largest diameter firebrand were used because results by Hadden et al. [2] and
Manzello et al. [18, 34] found that larger hot objects are more likely to produce
ignition.
Subsequently, tests under forced flow were conducted. These experiments were
conducted in front of the outlet of a laminar blower, the characteristics of which are
described in Singh and Gollner [44]. A series of tests was conducted in which the pile
mass was kept constant (100 g initial of 12.7 mm firebrands) and the wind velocity
was increased from 0.85 m/s to 1.98 m/s. These tests were used to used to identify
an ignition condition (wind velocity and pile size) which would reliably transition
to flaming ignition. The repeatable ignition condition for a flat sheet of OSB with
surface area 18 cm x 18 cm was 1.84 m/s wind velocity and 100 g initial mass of
36
12.7 mm firebrands. This wind speed is consistent with average wind speeds found
in fires (e.g. 1.8 m/s in the New Jersey Pine Barrens prescribed burns [45]).
A video camera recorded experiments at 29 fps (frames per second) from an
elevated side view, angled downward towards the experiments and allowing coverage
of the thin-skin calorimeter array to be determined. A FLIR ThermCAM SC3000
infrared camera (IR) camera with a spectral response of 8-9 µm recorded IR video
from directly above the set-up and recorded in the temperature range of 350-1500°C.
This range was chosen to cover the initial high temperatures of the firebrands. The
emissivity of the firebrands was unknown, though previous researchers have used
values between 0.6 - 1.0 [7]. A value of 0.92 was chosen; however, IR results were
meant to give a qualitative rather than quantitative view of heating from the top of
the pile.
For these tests, three wood configurations were compared: a flat board anal-
ogous to that used for heat flux tests, an L-shaped configuration with the vertical
board of the L perpendicular to the wind direction, and a crevice of width 5 mm
and depth 26 mm with the crevice perpendicular to the wind direction. These last
two configurations simulate the edge of a home where the deck meets the wall of
a house and decking boards, respectively. The crevice sizes correspond to typical
configurations [1]. Both of these configurations are known to be particularly suscep-
tible to firebrand accumulation [37]. All three ignition configurations with relation
to the wind direction can be seen in Figure 3.5.
Before testing, the mass of the OSB sample was taken so that total mass loss
could be calculated. Firebrands were deposited in the middle of the OSB sample
37
Figure 3.5: Schematic of recipient fuel configurations oriented with respect to wind
direction.
for both the flat and crevice samples, and in the corner of the L-shape, for the L-
shaped sample. Tests were continued until the IR camera registered no variation in
temperature between the sample and surrounding area. After testing finished, the
firebrands were removed from the OSB by tapping gently on the edge of the OSB,
and the final mass of the OSB sample was taken.
3.5.2 Thin-Skin Calorimeter Forced Flow Experiments
Heat flux tests using thin-skin calorimeters were also conducted under the
forced flow conditions found to produce ignition. Figure 3.6 shows the experimental
set-up for the wind tunnel experiments. The experimental board was the same board
used in the ambient tests. The procedures for checking the thin-skin calorimeters
before testing were the same. An additional step to prepare for experiments was
to focus the IR camera. In order to do this, a metal bracket was heated using
a butane blow torch and then placed on the experimental board beneath the IR
camera. The camera was then focused from the computer until the metal bracket
38
was clear. Otherwise, all procedures for the testing were the same. The wind tunnel
was turned on and set to the correct velocity at the same time as the data acquisition
was begun. Again tests were stopped when thin-skin calorimeter readings decreased
below 27°C.
Figure 3.6: Schematic side view of wind tunnel set-up for thin-skin calorimeter heat
flux tests, showing direction of wind and angles of cameras. Not drawn to scale.
39
Chapter 4: Measurement Techniques and Calibration Results
4.1 Overview
Two sets of experiments were described in the previous chapter: firebrand
pile tests using a water-cooled heat flux gauge and using thin-skin calorimeters and
ignition experiments. Key quantities were acquired for the different experiments:
temperature, heat flux, visual observations, and mass. Two measurement methods
were used to obtain temperatures: thermocouples and an IR camera. Likewise,
two measurement methods were used to acquire heat flux: thin-skin calorimeters
and a water-cooled heat flux gauge. During all tests using thin-skin calorimeters,
thermocouples also measured temperature of the firebrands. A load cell was used
for thin-skin calorimeter tests under ambient conditions. Video recording was taken
of all experiments,
4.2 Heat Flux: Water-Cooled Heat Flux Gauge
As a result of the uncertainties associated with the use of thin-skin calorime-
ters, a water-cooled heat flux gauge was also used to measure heat flux during these
experiments. The water-cooled heat flux gauge was a 1.27 cm (1/2 in) Medtherm
40
GTW-7-32-485A heat flux transducer. The water-cooled heat flux gauge had a lin-
ear calibration with voltage and was re-calibrated on a cone calorimeter against a
reference gauge. In preliminary experiments, a 2.54 cm Medtherm heat flux gauge
was also used; however, this gauge was not used during any experiments described
in the Results section.
4.2.1 Results of Gauge Size Comparison
Initially, a larger heat flux gauge of 2.54 cm diameter was used to conduct
preliminary tests. Due to a sharp drop-off in heat flux after an initial high peak, it
was determined that the water cooling of the heat flux gauge could potentially be
cooling the firebrand piles, particularly for smaller mass tests, such as those using
a single firebrand.
A comparison of 50 g tests (shown in Figure 4.1) using both the 2.54 cm and
1.27 cm water-cooled heat flux gauges show that the heat fluxes measured by the
larger gauge are all lower than the average heat flux of the two gauges, while the
smaller gauge recorded higher heat fluxes over the entire test length. It was inferred
that cooling at these high masses would indicate even more cooling at lower mass
piles. At longer times for larger piles (e.g. after 500 s), it is possible that some local
cooling could occur as the mass and temperature of the pile decrease; however, it is
expected that the cooling will be decreased with the smaller gauge.
41
Figure 4.1: Comparison of heat fluxes measured for a 50 g initial mass 12.7 mm
diameter firebrand pile.
4.3 Heat Flux: Thin-Skin Calorimeters
Thin-skin calorimeters were used to measure the spatial distribution of heat
flux beneath the firebrand piles. Because the thin-skin calorimeters were fabricated
and calibrated in-house, the following sections provide details on the fabrication,
calibration, and method of calculating heat flux from the temperatures measured
on the backside of the thin-skin calorimeters.
42
4.3.1 Fabrication
Thin-skin calorimeters were fabricated using 0.25 mm K-type wire and Inconel
alloy 625, a nickel-chromium alloy, used for its high thermal-fatigue strength and
melting temperature above 1290°C [47]. The thermocouple wires were welded to
the back surface of the calorimeter using a HOT SPOT TC spot welder used to
make thermocouples. Wires were spaced 1.6 mm apart in accordance with ASTM
E459 [46]. The area of the calorimeter surface was 1 cm2. It was important to
decrease the area of the thin-skin surface because the calculation of heat flux from
a thin-skin calorimeter uses the assumption of lumped capacitance. The thickness,
δ, of the Inconel was 0.508 mm. ASTM E459 [46] provides an equation to calculate
the optimum thickness for the calorimeter:
δopt =3
5
k(Tmax − T0)q
(4.1)
where k is the thermal conductivity of the thin-skin calorimeter in W/mK, Tmax
is the maximum temperature of the thin-skin calorimeter in K, T0 is the initial
temperature, and q is the heat flux in W/m2. An average thermal conductivity value
was taken by averaging the values of k at the maximum and initial temperatures
(values found in [47]). A maximum temperature of 760°C was chosen as a typical
average maximum value for large pile tests and an initial temperature of room
temperature, 21°C, was used. The k values for these temperatures are 20.8 W/m°C
and 9.8 W/m°C, respectively [47]. Results of tests using the water-cooled heat
flux gauge found maximum heat fluxes around 60 kW/m2. When these values are
43
substituted into Equation 4.1, a value of 0.113 m (113 mm) for δopt is found. The
δopt value is based on optimizing the maximum exposure time; however, because
the sensors continued to measure heat fluxes comparable to the water-cooled gauge
throughout experiments, it was assumed that exposure time was sufficient for this
application.
As the optimum thickness is unrealistic for the small-scale experiments, metals
of three thickness (1.27 mm, 0.8128 mm, 0.508 mm) were compared for accuracy
and time response. These thicknesses were chosen based on availability. Thin-
skin calorimeters of all thicknesses tested produced similar heat flux readings when
tested in a cone calorimeter at a set heat flux; however, the thicker calorimeters had
a slower time response, making the thinnest calorimeter more useful for transient
readings. Although the 0.508 mm thick calorimeters produced the shortest time
response, the thickness of the calorimeter is on the order of the diameter of the
thermocouple wire, introducing an additional potential error.
After fabrication, thin-skin calorimeters were exposed to a butane blow torch
flame for five minutes in order to tarnish the surface of the metal. Flame exposure
was conducted because the metal naturally became tarnished when the thin-skin
calorimeters were beneath the firebrand pile. Because the emissivity of the metal
was expected to change when it became tarnished, the metal was exposed before
testing in an attempt to keep the emissivity fairly constant between early and late
tests. One large source of uncertainty was the emissivity of the tarnished metal.
This uncertainty was accounted for by using a correction factor in the calculation
of heat flux, described in Section 4.3.3.
44
For preliminary tests, the surface of the thin-skin calorimeter was painted using
Zynolyte® Hi-Temp paint with a known emissivity of approximately 0.94 [48]. Due
to issues obtaining additional paint from the same manufacturer, this paint was
not available for the thin-skin calorimeters used in the majority of experiments.
Medtherm high temperature optical black coating, also of emissivity 0.94, was briefly
tested as a potential substitute; however, the paint underwent a reaction when in
contact with the firebrands, changing color and, thus, emissivity. As a result, the
tarnished thin-skins with unknown emissivity were used for all experiments described
in the Results section. Emissivity may change with temperature for this metal;
however, the temperature dependence is also unknown.
4.3.2 Calibration
The ASTM E459 standard [46] describes the use of thin-skin calorimeters in
radiative and convective environments. Even in these environments, calibration is
necessary to ensure that the calorimeters produce accurate heat fluxes. Hildalgo et
al., for example, show the initial discrepancies between a known incident radiant
heat flux and heat fluxes measured by a thin-skin calorimeter [38]. Uncertainties
regarding the heat flux readings arise out of the difficulty of not having a complete
knowledge of all of the components of the heat transfer processes. In particular, the
thin-skin calorimeter heat transfer is calculated with assumptions of lumped capac-
itance, one-dimensional heat transfer, known material properties, and a constant
emissivity [46]. In order to account for these uncertainties, thin-skin calorimeters
45
were calibrated using a radiant propane heater. A cone calorimeter was used once
as a comparison, but this method was not used extensively enough to provide a solid
calibration.
The calibration with the radiant heater required that the thin-skin calorimeter
be oriented in a vertical position, while the thin-skin calorimeters in the test array
were oriented horizontally. This difference has an effect on the convective losses from
the gauge. Although the Nusselt number correlation was changed for the vertical
and horizontal orientations, an additional calibration was conducted using a cone
calorimeter. Two thin-skin calorimeters were calibrated using this method. In this
calibration, a reference water-cooled heat flux gauge and the thin-skin calorimeter
to be calibrated where placed adjacent to one another beneath the center of the
cone. For one calibration, the thin-skin calorimeter was calibrated separately at
two point heat fluxes. For the other calibration, one thin-skin calorimeter was
calibrated at seven heat fluxes from 3-53 kW/m2, chosen to cover the range of heat
fluxes expected. In the first calibration, a painted thin-skin calorimeter was used;
however, the second calibration was for a tarnished thin-skin calorimeter, similar
to those used in the tests. Use of the radiant heater allowed a greater number of
gauges to be calibrated, thus providing a larger sample size for comparison.
There are still uncertainties associated with this calibration method. It uses
an incident radiant heat flux; however, the calorimeters are exposed to conductive
heat fluxes as well during the firebrand experiments. A method for conductive
calibration was not developed for comparison. Additionally, in the vertical set-up,
there is a significant convective component. On the other hand, in the experiment,
46
the assumption is made that, during firebrand pile tests, convective losses from
covered thin-skin calorimeters are negligible.
4.3.3 Theoretical Formulation
Rather than measuring heat flux, the thin-skin calorimeter measures temper-
ature, which can then be used to calculate heat flux. An equation for the total heat
flux from the firebrand pile to the thin-skin calorimeter can be found by conducting
an energy balance on a control volume of the thin-skin calorimeter, as shown in
Figure 4.2.
Figure 4.2: Energy balance around control volume of a single thin-skin calorimeter.
Estor = Ein − Eout (4.2)
where Ein denotes the energy flux into the control volume and Eout denotes the
energy flux out of the control volume. Using Equation 4.2 for a general energy
balance and the heat transfer terms from Figure 4.2 produces the heat transfer
balance shown in Equation 4.3.
47
q′′net − q′′conv − q′′rerad − q′′stor − q′′cond = 0 (4.3)
where q′′net is the net heat transferred into the thin-skin calorimeter from the firebrand
pile, q′′conv is the energy transferred away from the thin-skin calorimeter via natural
convective cooling, q′′rerad is heat reradiated from the thin-skin calorimeter to the
environment (or to the firebrand pile), q′′stor is stored heat, and q′′cond is heat conducted
through the thin-skin calorimeter to the wire and insulation, with all quantities
measured in W/m2.
Convective cooling occurs from the thin-skin calorimeter array to the environ-
ment via
q′′conv = h(TTS − T∞) (4.4)
where h is the convective heat transfer coefficient in W/m2K, TTS is the temperature
of the thin-skin calorimeter in K, and T∞ is the ambient air temperature in K.
The thin-skin calorimeter radiates heat following the Stefan-Boltzmann Law:
q′′rerad = εσ(T 4TS − T 4
∞) (4.5)
where ε is the emissivity of the Inconel metal and σ is the Stefan-Boltzmann con-
stant.
The heat storage rate is defined as
q′′stor = ρcpδdT
dt(4.6)
where ρ and cp are the density (in kg/m3) and specific heat (in J/kg K), respec-
tively, of the Inconel metal, both provided as a function of thin-skin calorimeter
temperature in [47], and δ is the thickness of the thin-skin calorimeter in m.
48
A correction term is calculated in place of the conductive heat transfer rate
to the surrounding insulation. This correction term is assumed to be a fraction
of the incident radiative heat flux as it should be small and both are assumed to
be temperature dependent. The correction thus also takes into account that the
emissivity of the tarnished metal is unknown, as is the conduction. The correction
heat rate is calculated as:
q′′cond = CαTS q′′rerad (4.7)
where C is the C-factor as a function of temperature found by calibration as de-
scribed in 4.3.2, αTS is the absorptivity of the Inconel metal.
Another method of calculating the correction factor is presented in Hildalgo,
et al. [38]:
q′′cond = CαTS q′′rerad (4.8)
where q”inc,rad is the incident radiant heat flux in W/m2 calculated as
q′′inc,rad =1
αTS(1− C)
[q′′stor + q′′rerad + q′′conv
](4.9)
following Equation 11 from [38]. In this case, the C-factor method of calculating
q′′cond is used because the conduction into the wires is unknown, the temperature of
the insulation board is unknown, as is the final emissivity of the Inconel metal.
When the expressions from Equations 4.4 - 4.8 are substituted into Equation
4.3 (with the expression for q”inc,rad substituted into the expression for q”cond, the
following expression for q”net can be found:
49
q′′net =1
1− C[q′′stor + q′′rerad + q′′conv
](4.10)
4.3.4 Results of Calibration
The purpose of the calibration was to find a correction factor which could
be applied to the heat flux calculations in the experimental use of the thin-skin
calorimeters. The C described in the previous sections was found following the
method of Hildalgo et al. where
C =αTS q
′′inc − q′′lossesαTS q′′inc
(4.11)
where q′′inc is the incident heat flux measured by the reference water-cooled heat flux
gauge. This term differs from q′′inc,rad, which is theoretically what a reference gauge
would measure during an experiment, whereas q′′inc is what the reference water-cooled
heat flux gauge measures during the calibration.
q′′losses = q′′stor + q′′rerad + q′′conv (4.12)
which are calculated as described in the Section 4.3.3, previously.
The correction factor was applied to the q′′cond term, which could better be
described as a correction term to take into account the uncertainties associated with
the emissivity of the Inconel when tarnished and the conduction into the insulation
and into the wires of the thin-skin calorimeter. A comparison of the corrected
50
net heat flux, q′′net, was plotted against q′′inc to determine the effectiveness of the
correction factor. An example of such a plot is shown in Figure 4.3.
Figure 4.3: Components of heat flux for one radiant heater thin-skin calorimeter
calibration test.
It is clear that the correction factor does not adequately correct to the full
incident heat flux in this environment. This difference is caused by the use of a
correction based on reradiation, rather than the incident heat flux. As a result,
the thin-skin calorimeter measurements are taken only as a qualitative indication
of trends; however, it is important to note that, even were the incident radiant
heat flux used to calculate the correction, the results would still only be qualitative
due to uncertainties that will be discussed in further sections. A comparison of
different C-factors found that the dependence on C-factor is so great that minor
changes have a large impact on the total correction. Due to the uncertainty in the
calibration, the heat flux calculated using the correction factor is expected to have
51
a high uncertainty. The problems and uncertainties associated with the calibration
will be discussed further in the Conclusions and Future Work, Chapter 8.
The correction factor was plotted as a function of temperature and fitted in
order to get C as a function of temperature to apply during the calculation of heat
flux in experiments. For the fit, the spike at the low and high temperatures were
disregarded. These spikes occurred when the calibration data recorded ambient
temperatures after the rest of the calibration was finished.
Figure 4.4: Linear fit to correction factor.
4.3.5 Application of Calibration to Experiments
Due to the uncertainties associated with the calibration method, the heat flux
values obtained using the thin-skin calorimeters have potentially high uncertainties.
52
Comparison with results from the water-cooled heat flux gauge is used to show
that, despite the uncertainty, heat flux values from the thin-skin calorimeters follow
similar trends. As a result, the important contribution of the thin-skin calorimeter
data is the illustration of qualitative trends and spatial heating.
The thin-skin calorimeters measured temperature as a function of time. A
plot of the time-dependent temperature for all sixteen thin-skins calorimeters in the
array can be seen in Figure 4.5. This figure and subsequent figures will be shown of
a single test of 12.7 mm diameter firebrands with a deposited pile mass of 9.6 g.
Using these temperatures, heat flux as a function of time was calculated for
each individual thin-skin calorimeter. The total heat flux was calculated as:
q′′net = q′′rerad + q′′stor + q′′cond (4.13)
where q′′rerad, q′′stor, and q′′cond are calculated as described in Section 4.3.3, and q′′cond is
the correction term using the C-factor found via calibration.
Convective losses were not considered for the firebrand pile experiments, except
during single brand experiments, as most sensors were blocked for part of the test.
Later in the test, as firebrands turned to ash, it is likely that convective cooling
occurred. Additionally, convective losses from the sensors were certainly present for
exposed sensors. Considering these losses negligible could potentially result in lower
net heat fluxes later in the test and for exposed thin-skin calorimeters. Figure 4.6
shows the raw heat flux calculated using the thin-skin temperatures for the same
test shown in the previous figure.
53
Figure 4.5: Thin-skin calorimeter temperature as a function of time for 16 thin-skins
in array. Thin-skin calorimeter locations can be found in Figure 3.3.
The water-cooled heat flux gauge results were used to help determine the
accuracy of the thin-skin calorimeter measurements. Figure 4.7 shows a plot of
the average heat fluxes measured using the water-cooled heat flux gauge for a test
of the same pile size and diameter firebrand as shown previously for the thin-skin
calorimeters. Figure 4.8 shows a magnified version of Figure 4.7 to capture the
long-time behavior.
54
Figure 4.6: Total heat flux imparted to the thin-skin calorimeters, calculated using
measured thin-skin calorimeter temperatures.
In order to compare thin-skin calorimeter heat flux results with those obtained
using the water-cooled heat flux gauge, time averages were taken of the covered thin-
skins for each test. A further description of this procedure can be found in Chapter
5, Section 5.4. Figure 4.9 shows this comparison between heat flux measured by the
thin-skin calorimeters and heat flux measured by the water-cooled heat flux gauge.
The averages are taken over five test repetitions for the water-cooled heat flux gauge
and over nineteen test repetitions for the thin-skin calorimeters in this particular
configuration.
55
Figure 4.7: Averaged heat flux results for a 12.7 mm diameter firebrand, 9.6 g
deposited mass pile, measured using the water-cooled heat flux gauge.
Figure 4.9 shows a representative trend comparison between the two heat flux
measurement methods that can be seen in all pile sizes except the single brand.
The water-cooled gauge peaks early, then drops to a steady value, before decreasing
again at the end of the test as the firebrands cool. The thin-skin calorimeters peak
later in the test, which is consistent with the slower time response. They sustain a
higher heat flux value than the water-cooled heat flux gauge for the rest of the test.
The lower heat flux from the water-cooled heat flux gauge could be consistent
with potential cooling or the difference in heat transfer between a cold object (the
gauge) and hot object, as opposed to the thin-skin calorimeters, which increase in
56
Figure 4.8: Magnified replication of previous figure of averaged heat flux results
for a 12.7 mm diameter firebrand, 9.6 g deposited mass pile, measured using the
water-cooled heat flux gauge, showing long-time trends.
temperature throughout the duration of the test. Despite the problems with the
calibration, the thin-skin calorimeters read heat fluxes broadly comparable to those
measured using the water-cooled heat flux gauge.
57
Figure 4.9: Plot of averaged heat flux over all repetitions for a 12.7 mm, 9.6 g test
to compare heat fluxes obtained using thin-skin calorimeters and the water-cooled
heat flux gauge.
58
Chapter 5: Results of Firebrand Pile Tests
5.1 Overview
Several types of data were acquired throughout the different categories of tests,
as described in Chapter 4. Firebrand pile tests under ambient conditions were con-
ducted using both thin-skin calorimeters and a water-cooled heat flux gauge. The
main results obtained for both sets of tests are heat flux versus time curves. Results
obtained using the thin-skin calorimeter array also included temperature measure-
ments from the thermocouple and thin-skin calorimeter and spatial distributions of
heat flux over the entire array. Initial time series heat flux curves were presented
at the end of Chapter 4. Results in this chapter will include visual observations,
the thin-skin calorimeter and thermocouple temperatures, components of heat flux
measured by thin-skin calorimeters, spatial distributions of heat flux, and methods
of comparing heating from piles.
5.2 Visual Observations
Visual observations provided interesting information regarding the burning
of firebrand piles. Although all of the firebrands were fabricated together, there
59
were differences between firebrands even when they were first deposited on the
test board. Some firebrands appeared to be mostly charred, while others glowed
completely throughout the entire brand. The most common behavior was glowing
at both tips of the firebrand. The varieties of glowing can be seen in Figure 5.1,
which shows single firebrands and firebrand piles immediately after deposition on
the experimental board.
Figure 5.1: Image sequence showing varying firebrand pile sizes, increasing in diam-
eter from left to right and increasing in pile mass from top to bottom.
Though firebrands were at first deposited in a glowing state, they quickly
transitioned to ash in many tests. The beginnings of this transition can be seen in
Figure 6.2 in the following chapter. In many cases, the entire top of the pile would
60
turn to ash, which can be seen in Figure 6.4 in the following chapter. It was assumed
that the ash indicated that the test would be coming to an end; however, the ash
was later thought to insulate hotter glowing cores of firebrands. This phenomenon
was especially apparent in wind tunnel tests. At times the wind blew away sections
of ashy material revealing glowing firebrands beneath. This behavior may explain
differences in temperature measured by the IR camera, a surface thermocouple, and
the thin-skin calorimeters.
5.3 Thermocouple and Thin-Skin Calorimeter Temperature Results
Both the thin-skin calorimeters and a thermocouple measured temperatures
during the thin-skin calorimeter tests. The temperatures measured by these two
instruments are compared in order to determine whether the thin-skin calorimeters
can provide a realistic spatial distribution of temperature beneath the firebrand pile.
Figures 5.2–5.4 show three methods of comparing the temperatures obtained
by the thermocouple and the thin-skin calorimeters. The figures show a 9.6 g de-
posited mass pile of 12.7 mm diameter firebrands under ambient conditions. Figure
5.2 shows the temperatures from all sixteen thin-skin calorimeters alongside the
temperature from the thermocouple as a function of time.
It is clear from Figure 5.2 that the thermocouple experiences a much faster
time response than the thin-skin calorimeters. This result is expected, as the time
response of the thermocouple should be approximately 10 s, based on the wire
gauge, and the response of the thin-skin calorimeters was found to be on the order
61
Figure 5.2: Temperature as a function of time as measured by thin-skin calorimeters
and thermocouple beneath 12.7 mm diameter firebrand pile of 9.6 g deposited mass.
of 100–150 s. The thin-skin calorimeters also continue reading high temperatures
for a much longer period of time. There are two potential reasons for this behavior.
First, the thin-skin calorimeters are larger and thus store more heat. Second, as
different parts of the firebrand pile heat and cool, the thermocouple may not always
be in the optimum location to measure the pile temperature. It is relevant to note
that peak temperatures from the thin-skin calorimeters and from the thermocouple
are similar.
In order to obtain a clearer picture of the heating in the pile, the thermocouple
temperature was plotted with the temperatures of the four surrounding thin-skin
62
calorimeters. These results are shown in Figure 5.3. This figure shows similar
trends to the previous plot (Figure 5.2) of all thin-skin calorimeter temperatures;
however, in this case, two of the thin-skin calorimeters nearest to the thermocouple
follow almost the same temperature trend as the thermocouple itself. The thin-skin
calorimeters with this general trend do not capture the same peak temperature, but
capture the same cooling behavior.
Figure 5.3: Temperature as a function of time as measured by the thermocouple and
its four immediate nearest thin-skin calorimeters for a 12.7 mm diameter firebrand
pile of 9.6 g deposited mass.
Figure 5.4 shows a further comparison of temperatures using the averaged tem-
perature for all covered thin-skin calorimeters in this test. Although peak tempera-
63
tures are not captured, the thin-skin calorimeters capture the general temperature
trend. This trend suggests that thin-skin calorimeters may function as a relatively
accurate representation of the temperature beneath a firebrand pile at long times.
As a matter of fact, the thin-skin calorimeters could provide more accurate surface
temperatures than they do heat fluxes.
Figure 5.4: A comparison of temperature as a function of time for the thermocouple
and an average temperature of the thin-skin calorimeters initially covered by the
firebrand pile for a 12.7 mm diameter firebrand pile of 9.6 g deposited mass.
64
5.4 Heat Flux Results
Different aspects of the heat flux measurements show important trends. Using
the thin-skin calorimeter data, we can see the relative magnitude of the different heat
transfer components. The thin-skin calorimeters also provide a spatial distribution
of heat flux as a function of time throughout the test. Finally, comparisons can be
drawn between the methods of measuring heat flux.
5.4.1 Components of Heat Transfer
In order to calculate net heat flux imparted to each individual thin-skin calorime-
ter, individual components of heat flux were calculated as functions of time and
summed. Figure 5.5 shows the components of heat transfer for a single thin-skin
calorimeter central to a large firebrand pile. These values give an example of the
approximate magnitudes of heat flux components calculated for piles. Re-radiation
plays an important role in the heating, which is consistent with the high tempera-
tures and glowing observed. Unfortunately, the conduction correction term domi-
nates, indicating that the thin-skin calorimeters can be subject to large errors given
the high dependence of the correction term on individual heat transfer components
(e.g. radiation). This calibration may then compound uncertainty in the emissivity
value.
Convection losses were not included when calculating the net heat flux for
piled configurations as most sensors were covered; however, convection was added
to the calculations for a single firebrand, as this configuration left most of the thin-
65
Figure 5.5: Components of heat transfer for a single thin-skin calorimeter during a
piled test.
skin array exposed. Figure 5.6 shows a similar plot of heat flux components for a
test using a single 12.7 mm firebrand. Here convective losses form the main portion
of the heat transfer, though re-radiation is also important.
66
Figure 5.6: Components of heat transfer for the covered thin-skin calorimeter of a
single firebrand test.
5.4.2 Spatial Distribution of Heat Flux
Figure 4.6 in the previous chapter shows a wide distribution of heat flux behav-
iors for different thin-skin calorimeters. Some thin-skin calorimeters reach a peak
heat flux and drop off quickly, while others sustain a peak heat flux for over 500 s.
Some calorimeters also peak later in the test. In order to understand these trends,
video data was analyzed to determine which thin-skins were completely covered by
firebrands at the beginning of the test, which were partially covered by firebrands,
and which were completely exposed or not covered at all. Figure 4.6 was re-plotted
67
as Figure 5.7 showing which thin-skin calorimeters fell into each of these categories
of coverage.
Figure 5.7: Heat flux as a function of time for a 12.7 mm diameter, 9.6 g deposited
mass test, measured using the thin-skin calorimeter array with coverage of thin-skins
denoted.
The majority of the array is covered at the beginning of this test; however,
similar plots of tests with smaller firebrand masses show different proportions of
coverage. In order to compare the effect of coverage, as well as compare curves be-
tween plots, averages were taken of temperatures and heat fluxes based on the three
coverage levels identified. These averages can be seen for this test for temperature
and heat flux, in Figure 5.8 and Figure 5.9, respectively.
68
Figure 5.8: Averaged temperatures of thin-skin calorimeters based on coverage: full,
partial, or no coverage (12.7 mm diameter, 9.6 g deposited mass test).
Figures 5.8 and 5.9 use the coverage from the firebrand pile of the thin-skin
calorimeter array at the beginning of the test. This method produces trends with
how long different heat fluxes last.
Another way to visualize these trends is by looking at the spatial distribution
of heat flux. The thin-skin calorimeter array provides a way to look at how heating
changes based on pile location. Figure 5.10 shows how heat flux evolves spatially
throughout a single test. The areas of highest heat flux change throughout the test.
The first time step shows two areas beginning to heat which are further heated in
the following time step. At the third time step, though, the area of highest heat
69
Figure 5.9: Averaged heat flux of thin-skin calorimeters based on coverage: full,
partial, or no coverage (12.7 mm diameter, 9.6 g deposited mass test).
flux has shifted towards the middle of the array. The middle area continues to have
the highest heat flux for the rest of the test. The dashed lines represent 14 kW/m2,
a critical heat flux condition for radiant ignition of wood.
Figure 5.11 shows three spatial maps of heat flux for a single firebrand and for
a pile of 9.6 g of firebrands. The comparison shows an instantaneous snapshot of
heat flux, the average heat flux throughout the test, and the maximum heat flux for
the entire test. The single firebrand heats a large area, indicating the importance
of re-radiation from the firebrand. The heating in the large pile does not show the
same trends. Similar to Figure 5.10, the heated area changes. For these figures, the
color scale changes between figures to show the heated area trends optimally.
70
Figure 5.10: Evolution of spatial distribution of heat flux for a 5 g deposited mass
pile of 12.7 mm at 100 s, 150 s, 400 s, 900 s, and 1400 s after firebrands are deposited.
71
Figure 5.11: Comparison of spatial heat flux maps for (left) a single firebrand, and
(right) the largest pile size, 9.6 g deposited mass. From top to bottom, heat flux
maps shown an instantaneous heat flux approximately 1 min into the test, averaged
heat flux over the entire test, and maximum heat flux for each location. Due to
large heat flux differences, the color bar values change for each plot.72
5.5 A Comparison of Heating From Piles
Two main quantities of interest were varied during firebrand pile experiments
to determine what parameters affected heat flux values the most: firebrand diameter
and pile size (measured as mass – initial or deposited – and as number of firebrands
per pile). The following sections shows the comparison of heat flux curves as a
function of firebrand diameter and deposited pile mass. Subsequent sections will
compare point values for these tests as a function of diameter and pile mass.
5.5.1 Effect of Brand Diameter
Firebrand size was found to be important in previous studies using a single
or a very small number of firebrands [18, 34]. Throughout these tests, firebrand
diameter was varied to determine whether it was an important parameter affecting
heat flux. Figures 5.12 through 5.15 show results of comparing diameters for a single
pile size. These results were plotted using the average heat flux curve over all of the
repetitions of a given test measured using the water-cooled heat flux gauge.
A slight variation (up to 0.2 g) between pile sizes can be noted; however, uncer-
tainty in pile mass due to load cell readings and differences between individual tests
make the masses shown comparable within the uncertainty of mass measurements.
For these comparisons, the pile size used is the deposited rather than initial mass;
however, the experimental matrix was produced using initial masses. Since differ-
ent diameter firebrands lose different proportions of mass when burning, a similar
initial mass does not necessarily translate to a similar deposited mass. Deposited
73
mass is used due to trends found between heating and deposited mass, which will
be discussed in Section 5.5.3.
For a single firebrand (Figure 5.12), the larger diameter produced a higher heat
flux over time and continued heating for longer than the smaller diameter did. This
result is consistent with literature results that found that a single larger firebrand
could initiate ignition in a porous fuel bed when a smaller firebrand could not [18].
Figure 5.12: A comparison of averaged point measurements from the water-cooled
heat flux gauge of 6.35 mm and 9.52 mm diameter firebrands for a single firebrand.
The larger diameter produces a higher heat flux and lasts longer.
74
Figure 5.13: A comparison of averaged point measurements from the water-cooled
heat flux gauge of 6.35 mm and 9.52 mm diameter firebrands for a 1 g pile, deposited
mass. Differences in heat flux are well within the standard deviation of the average.
75
Figure 5.14: A comparison of averaged point measurements from the water-cooled
heat flux gauge of 6.35 mm and 12.7 mm diameter firebrands for a 2.8 g pile,
deposited mass. There is no trend apparent between heat flux and diameter.
For the largest pile compared here (Figure 5.15), the larger diameter firebrands
produce higher heat fluxes; however, this trend does not hold for the medium size
piles (Figures 5.13 and 5.14). For these piles, the differences are well within the
standard deviation of the test. Despite a possible trend for the large piles, there is
not sufficient difference to indicate that larger diameters produce higher heat fluxes.
It is possible that the contact from individual firebrands becomes less important in
a pile where re-radiation within the pile could provide more heating to a recipient
fuel. Section 5.5.2 will show that, while diameter does not produce a strong trend,
deposited mass appears to show a better trend with heating.
76
Figure 5.15: A comparison of averaged point measurements from the water-cooled
heat flux gauge of 9.52 mm and 12.7 mm diameter brands for a 5 g pile, deposited
mass. The drop-off of the 9.52 mm curve is due to averaging; the average heat flux
curve was calculated only as long as the shortest test. The larger diameter produces
higher heat fluxes.
5.5.2 Effect of Pile Mass
Larger masses of firebrands produced raised temperatures and heat fluxes for
a longer period of time. This fact was observed qualitatively during testing, but
shown quantitatively by comparing experiments using the same firebrand diameter,
but different masses. Figures 5.16-5.19 show heat flux as a function of time for
77
averaged curves of the four pile masses tested, for 6.35 mm, 9.52 mm, and 12.7 mm
firebrand diameters, respectively.
Figure 5.16: A comparison of averaged heat flux curves obtained using the water-
cooled heat flux gauge as deposited pile mass increases for 6.35 mm firebrand piles.
In all cases, the the steady heat flux value after the peak increased with increas-
ing pile size. It should also be noted that the largest pile size for each diameter had
the longest semi-steady period of heat flux. Tests with higher masses also produced
heating longer than tests with smaller masses. The averages do not always end at
ambient conditions. This cut-off is a result of averaging over all test repetitions –
some tests lasted longer than others, even at the same diameter.
78
Figure 5.17: A comparison of averaged heat flux curves obtained using the water-
cooled heat flux gauge as deposited pile mass increases for 9.52 mm firebrand piles.
Figure 5.18: A comparison of averaged heat flux curves obtained using the water-
cooled heat flux gauge as deposited pile mass increases for 12.7 mm firebrand piles.
79
Figure 5.19: A comparison of averaged heat flux curves obtained using the thin-skin
calorimeters as deposited pile mass increases for 12.7 mm firebrand piles.
5.5.3 Peak Heat Flux and Net Heating
Initial results provided heat fluxes as a function of time; however, it would be
helpful to have a parameter which could be used to quantify and compare the overall
heating between different tests. Such a parameter would be particularly helpful in
smoothing out the different shapes of the heat flux curves for the water-cooled heat
flux gauge and the thin-skin calorimeter array. Initially, the peak heat flux was used
as a comparison between tests. The peak heat flux was plotted against firebrand
diameter and deposited mass. Figures 5.20–5.23 show these results for both thin-skin
calorimeters and the water-cooled heat flux gauge.
80
Figure 5.20: Peak heat flux as a function of firebrand diameter from thin-skin
calorimeter tests.
81
Figure 5.21: Peak heat flux as a function of firebrand diameter from water-cooled
heat flux gauge tests.
82
Figure 5.22: Peak heat flux as a function of deposited mass from thin-skin calorime-
ter tests.
Plotting the peak heat flux as a function of either firebrand diameter or mass
deposited does not produce informative results. The peak heat flux increases when
a pile, rather than a single firebrand, is deposited; however, the values of peak
heat flux for piles do not vary significantly, and the variation is typically within the
standard deviation for a given diameter. This trend is particularly clear in Figure
5.23.
As the peak heat flux does not produce the most telling results, a net heating
parameter was introduced to compare tests. The net heating parameter was calcu-
lated as the area under the heat flux versus time curve and represents the total heat
83
Figure 5.23: Peak heat flux as a function of deposited mass from water-cooled heat
flux gauge tests.
imparted from the firebrand pile to the heat flux gauge throughout the test. The
results of plotting the net heating parameter as a function of diameter and deposited
pile mass are shown in Figures 5.24-5.26. Figure 5.26 shows that plotting the net
heating parameter as a function of pile mass results in a linear relationship, where
the total heat imparted by the firebrand pile increases as the pile size increases.
This result is expected as a larger mass of firebrands has a larger potential chemical
energy to release over time.
There is a question about an appropriate way to measure firebrand pile size.
Results using mass deposited have been used in the figures so far shown; however,
measuring the number of firebrands in a pile has been suggested as a potential
84
Figure 5.24: Net heating parameter as a function of firebrand diameter for water-
cooled heat flux gauge tests.
option. The net heating parameter from the water-cooled heat flux gauge tests was
plotted against the total number of firebrands in a pile in order to compare these
two methods of measuring pile size. Figure 5.27 shows net heating as a function of
number of firebrands in a pile.
Despite the fact that Figure 5.27 shows an approximately linear relationship
between number of firebrands and net heat released for each diameter, the trend
clearly varies with diameter. Figure 5.26 shows that the different trends of the
diameters collapses towards a single linear trend when plotted as a function of
deposited mass. These results indicate that mass may be a better metric than
firebrand number to describe pile sizes. As the number of firebrands for a given
85
Figure 5.25: Net heating parameter as a function of deposited pile mass for thin-skin
calorimeter tests.
mass pile varies significantly with brand diameter, it is unsurprising that pile mass
provides a better metric. For example, 37 wooden dowel pieces of 12.7 mm diameter
were required to produce 70 g of wood, whereas 142 pieces of 6.35 mm diameter were
required to produce the same initial mass.
Although the net heating parameter provides some potentially useful informa-
tion, it is calculated over an entire test. Taking the heating over an entire tests does
not take into account the fact that a critical condition (e.g. heat flux) is needed
to ignite a material. Previous experiments of both flaming and smoldering igni-
tion [10, 22, 23] have found that it is necessary to maintain a critical heat flux (or
temperature) for a given time in order to ignite a material. The time to ignition
86
Figure 5.26: Net heating parameter as a function of deposited pile mass for water-
cooled heat flux gauge tests.
is dependent on the heat flux and time to ignition decreases as heat flux increases;
however, below a critical heat flux, ignition will not occur.
87
Figure 5.27: Net heating parameter as a function of number of firebrands in the pile.
Each marker is an average of all repetitions conducted for a specific test condition.
Error bars are the standard deviation of the repetitions.
88
Chapter 6: Results of Ignition Tests
6.1 Overview
The following chapter describes the results of a series of ignition tests. Two
sets of tests were completed: tests using OSB as a recipient fuel using two firebrand
pile sizes and a set wind velocity. All initial tests using OSB found smoldering, so
flaming ignition was sought. A second set of tests replicated the pile size and wind
velocity, but were conducted on an inert surface so that heat flux and temperature
measurements beneath the pile could be made. Pile sizes were approximately 5 g
and 10 g deposited mass of 12.7 mm firebrands. The wind velocity was 1.84 m/s.
6.2 Visual Observations
When firebrand piles were initially deposited onto the recipient fuel, immediate
flaming occurred briefly, but was not sustained. This flaming occurred when the
firebrands were exposed to the external flow, and thus flames were sometimes present
upon deposition. This occurred for tests both with and without the recipient fuel.
Figure 6.1 shows this phenomenon – a small flame is visible on the top right hand
side of the image.
89
Figure 6.1: Side view of a flat ignition experiment immediately after firebrand pile
is deposited. A small flame occurs in the upper right hand portion of the image.
Wind direction right to left.
Although this initial flaming occurs, the flames last very briefly. Thus it is
not supposed that the flames from the firebrands directly cause ignition of dense
thermally thick fuels. Additionally, the immediate deposition of the entire firebrand
pile is unlikely to be found in a WUI fire. It is more likely that single firebrands
would gradually deposit on a pile over time.
Within one minute of the beginning of the test, flaming is visible on the sides
of the firebrand pile. Approximately a minute and a half into the test, a flame
anchored to the recipient fuel is visible, as shown in Figure 6.2. A visual sign of
potential smoldering is a spreading char front in areas not directly in contact with
firebrands. This charring could, possibly, also be the effect of high radiant heat
fluxes from the pile; however, in this case, blue flames are visible on the bottom
left hand corner of the pile and on the top left hand corner in the image. A bright
90
yellow flame is also visible anchored to the OSB directly to the right of the glowing
core of the firebrand pile.
Figure 6.2: Side view of flat ignition experiment over OSB 1 min into experiment.
Flames anchored to the fuel surface are visible. The wind direction is from right to
left.
As the experiment progresses, flaming continues from several locations on the
surface of the fuel on the side of the firebrand pile. Flaming of the glowing core
of the firebrand pile also occurs intermittently. Flaming of the fuel is sustained for
approximately 10 min. Figure 6.3 shows the same test as the previous figures at the
10 min mark.
After about 15 min, the test ceases flaming and the firebrand pile is blown
over the test section (see Figure 6.4). This likely occurs as the firebrand pile has
mostly turned to ash, becoming light enough to be lofted by the wind. The fuel
continues to smolder for nearly an hour with intermittent reheating. This test lasted
approximately an hour. Other flat tests followed the same approximate trends.
91
Figure 6.3: Side view of a flat ignition experiment over OSB 10 min into the exper-
iment. Flaming combustion is sustained for the recipient fuel. The wind direction
is from right to left.
6.2.1 Description of the Ignition Process
Initially, three possible processes for ignition were hypothesized. Figure 6.5
shows a conceptualization of possible processes. The first possible process identified
is that glowing or smoldering firebrands could transition to flaming and the flaming
firebrands could ignite the recipient fuel. The second possible process is that the
smoldering firebrands could heat the recipient fuel directly and cause the recipient
fuel to begin smoldering. In this case, it is supposed that the recipient fuel would
itself transition to flaming, possibly in front of the firebrands where there is more
oxygen available. The third possible process would be a long-term heating of the
recipient fuel, which would later transition to flaming ignition somewhere over the
surface of the fuel.
92
Figure 6.4: Side view of flat ignition experiment over OSB 15 min into the experi-
ment. Flaming combustion has ceased. The wind direction is from right to left.
Figure 6.5: Conceptualization of possible ignition processes that occur when fire-
brands ignite a recipient fuel.
Based on visual and IR observations of the tests, it appears that some form
of the second process described is most likely. The first process does not seem to
fit the behavior, as sustained flaming of firebrands did not occur for tests without a
recipient fuel, including those tests conducted under the exact same wind and pile
size configurations. On the other hand, the recipient fuel ignited fairly early in the
test, indicating that long-term heating before ignition may not accurately represent
the ignition process in this experimental scenario.
93
Figure 6.2 shows flaming ignition of the recipient fuel in two locations. On
the close edge of the OSB, blue flames are visible over a charred section of fuel.
Towards the center of the firebrand pile, a charred section of wood is visible in
front of the brightly glowing firebrands, and a flamelet appears to be anchored in
this location. Based on this description, it appears recipient fuel has been heated
directly. Flaming of firebrands for brief periods may have acted as a pilot ignition
source for the potentially smoldering fuel.
6.3 Thermal Characteristics at Ignition
For each ignition test, IR video was recorded in order to provide a qualita-
tive picture of the changing surface temperature. The video was used mostly for
qualitative purposes because of the unknown emissivity of the firebrands. IR video
was also recorded during tests with a nearly adiabatic surface used to measure heat
fluxes leading up to ignition. Other thermal characteristics obtained for ignition
experiments included the temperature beneath the firebrand pile, measured using a
thermocouple, and the thin-skin calorimeter array measuring the heat flux beneath
the pile.
6.3.1 Temperatures
Frames from the IR video were obtained for the key times described in the
visual observations section previously. They are shown in Figure 6.6 next to the
94
corresponding visual image. The top row shows some heating surrounding the fire-
brand pile, but the highest temperature is in the firebrand pile.
Figure 6.6: Visual image (left) and IR image (right) captures for, from top to
bottom: 1 min, 10 min, and 15 min into an ignition test. Note the changed scales
on the IR images. The wind is from the right side of images.
95
The middle image shows the ignited fuel at the back of the firebrand pile. The
bottom image shows the high temperatures of remaining firebrands, despite an ashy
pile with no visible combustion.
Thermocouple and thin-skin calorimeter temperatures were also measured dur-
ing these tests and are plotted in Figures 6.7 and 6.8. Figure 6.7 shows the overall
averaged temperature for covered thin-skin calorimeters, and Figure 6.8 shows the
four thin-skin calorimeters immediately nearest the thermocouple.
Figure 6.7: A comparison of the thermocouple temperatures and thin-skin temper-
atures averaged over all covered thin-skins for one test of 10 g deposited mass.
Both the thin-skin calorimeters and the thermocouple reach higher temper-
atures than the IR camera, which is perhaps reasonable given the uncertainty of
the emissivity. It is also possible heat losses from the surface and the insulating
96
Figure 6.8: A comparison of the thermocouple temperature with the temperatures
measured by the four thin-skin calorimeters surrounding the thermocouple for the
same test as previous figure.
effect of a pile of brands increases temperatures below the pile. The temperatures of
the thin-skin calorimeters remain higher longer than both the thermocouple and IR
camera, which is expected due to heat storage. The total time of raised temperatures
underneath the pile is approximately 40 min.
6.3.2 Heat Flux
Heat flux was measured under piles of 5 g and 10 g deposited mass under
the 1.84 m/s external airflow. Figure 6.9 shows the heat flux measured from all
97
sixteen thin-skin calorimeters for a single 10 g test. Heat flux results presented in
this section are for the same test presented in the previous section on temperature.
Figure 6.9: Heat flux as a function of time for the full thin-skin calorimeter array
for a 10 g deposited mass test with applied wind.
One aspect that heat flux measurements highlight is the spatial difference in
heat fluxes. In Figure 6.9, most of the thin-skin calorimeters peak around 500 s;
however, two thin-skin calorimeters (TSC 1 and TSC 5) peak more than 500 s later.
This difference in timing of the heat flux peaks may be a result of areas of the
firebrand pile reheating and the wind blowing the pile as more of it turns to ash.
Figure 6.10 shows the average heat flux as a function of time averaged for six
tests at 10 g deposited mass. There is high variation in the measured heat flux using
the thin-skin calorimeters.
98
Figure 6.10: Averaged heat flux curves for 10 g deposited mass tests with wind.
6.3.3 Spatial Distribution of Heat Flux
One phenomenon noticed in these tests was a reheating in different parts of
the firebrand pile. This trend was obvious in the IR video temperatures and the
time-dependent heat flux curves. The following series of spatial distributions of
the heat flux for a single test show how heat flux develops in different parts of the
firebrand pile.
Figure 6.11 shows the spatial distribution of heat flux approximately a minute
into one test, when ignition is expected. Figure 6.12 shows the same test over
ten minutes into the test. In this case, there is no recipient fuel. Nonetheless,
the location of heating has changed, as has the total heat. A different hot spot
has developed here. Note that the scales are with respect to the maximum and
99
minimum heat flux measured at a given time. The dashed line represents 14 kW/m2,
a minimum threshold for radiant ignition of wood.
Figure 6.11: Spatial distribution of heat flux approximately 1 min into a test with
10 g deposited pile mass. Wind direction right to left.
Figures 6.13 and 6.14 show the average and maximum heat fluxes obtained
throughout the entire test. The average is taken over the first two thirds of the test
to more accurately represent averages during higher active heating. It is interesting
to note that the locations of the highest heat fluxes do not necessarily correlate with
the locations of overall highest average heat flux.
100
Figure 6.12: Spatial distribution of heat flux over 10 min into the same test with 10
g deposited pile mass. Wind direction right to left.
Figure 6.13: Spatial distribution of heat flux averaged over the first two thirds of
the forced flow test with a 10 g firebrand mass.
101
Figure 6.14: Spatial distribution of the maximum heat flux obtained for each point
during the forced flow 10 g mass test.
102
Chapter 7: Discussion
7.1 Firebrand Pile Heat Flux Tests
An initial comparison of heat flux curves found using the thin-skin calorimeters
and the water-cooled heat flux gauge shows some clear trends. The shape of the
latter heat flux curves includes a sharp spike, which drops off quickly and then
steadies, sometimes only briefly, at a much lower value. On the other hand, heat
fluxes from the thin-skin calorimeters did not register a similar peak, but peaked
later and recorded higher heat flux values for a longer time.
There are definite limitations to both types of heat flux measurement meth-
ods used for these tests. The water-cooled gauge may be cooling firebrands, while
the calibration of the thin-skin calorimeters has such high uncertainties that the
values cannot be taken as accurate quantitative measures. Ultimately, an improved
method is needed to measure heat flux if quantitative values, not just trends, are
desired; however, trends provide important information on the parameters which
influence heating. One possible method would be a high temperature thermopile.
A thermopile was not used in this study due to concerns about durability at high
temperatures; however, the temperatures measured in these tests (up to 900°C) are
below the maximum temperature for one thermopile which was considered (up to
103
1000°C). There would still be potential issues with a thermopile, but this option
may be worth exploring.
The firebrand pile results also found that heat flux was not highly dependent
on firebrand diameter for large piles. Previous studies have found that small piles
and single firebrands are dependent on size; however, the ignition tests here show the
importance of large piles. It would be better to get a wider array of data for a single
firebrand diameter or size first. Varying pile mass is a more important parameter
and affects heat flux and net heating from tests. This knowledge is critical when
determining which parameters are most important to measure in order to estimate
a firebrand “flux” which is representative of exposure in WUI fire conditions. The
mass of smoldering firebrands deposited in a location, thus far, appears to be the
most important parameter.
7.2 Ignition Tests
Ignition tests in the wind tunnel indicated a potential process governing igni-
tion of a recipient fuel from a firebrand pile. This process of ignition highlights the
impact of a large mass of heated objects, rather than a single heated object. The
re-radiation within the pile also plays a key role, as does reheating. Reheating may
be an important parameter to consider in future ignition models, as it resulted in
re-initiation of flaming during some tests.
The applied airflow is clearly important as it produced ignition and higher heat
fluxes. Although these heat flux values may not be entirely accurate, comparing the
104
heat fluxes found for ambient and forced flow tests shows a drastic difference (see
Figure 7.1). Heat fluxes are seen to peak much higher under forced flow conditions.
This is expected to occur as increased airflow induces more oxidation which results
in higher heat release and higher temperatures of the brands, ultimately resulting
in higher surface temperatures at the text section and higher estimated heat fluxes
by thin-skin calorimeters. This heating, however decays much faster than under
ambient conditions for similarly-sized tests.
A similar trend can be seen with the net heating parameter when ambient and
forced flow tests are compared, particularly for the larger mass pile (Figure 7.2).
The difference appears to be that, in the wind-driven tests, the heat is transported
over a much smaller time, resulting in higher heat fluxes than with the ambient tests.
During ambient tests relatively the same amount of heat is transferred, except that
it is transferred over a longer time, resulting in lower heat fluxes.
105
Figure 7.1: A comparison of averaged heat flux curves, measured with thin-skin
calorimeters, under ambient and forced flow conditions. The dashed lines represent
initial mass 50 g test while the solid lines represent initial mass 100 g tests. All tests
use 12.7 mm diameter firebrands.
106
Figure 7.2: Net heating parameter as a function of deposited pile mass for ambient
and forced flow experiments using 12.7 mm diameter firebrands.
107
Chapter 8: Conclusions
8.1 Overview
A methodology for conducting heat flux experiments for firebrand piles has
been presented. A reliable method for producing piles without wind has been
developed. Heat fluxes from a firebrand pile were measured using both a water-
cooled heat flux gauge and an array of thin-skin calorimeters. Heat flux results were
connected to an ignition condition and the thermal characteristics of the ignition
configuration were described.
8.2 Conclusions
One of the most important aspects that this study highlighted is the diffi-
culty of reliably measuring heat flux from a firebrand pile. Attempts were made to
measure heat flux using both a water-cooled heat flux gauge and an array of thin-
skin calorimeters; however, these measurements had high uncertainties. They serve
better to illustrate trends and the approximate order of the heat flux, rather than
exact values. One of the difficulties with measuring heat flux from the thin-skin
108
calorimeters was the difficulty of applying the calibration due to the dominance of
different components of heat transfer in the calibration and in the experiment itself.
Several trends were identified which can motivate decisions on variables to
study in future work, both in the laboratory and the in the field. Pile size (mass) was
found to be the most important variable affecting heat flux curves and net heating.
The results from piled firebrands are also distinctly different than those from a single
firebrand, as the re-radiation within the pile appears to play an important role in
heating.
Under ambient conditions, diameter was found to have a limited influence on
the heat fluxes measured and net heating from firebrand piles, although it is known
that firebrand size can be a critical ignition factor for a single firebrand. Ultimately,
then, the diameter used for ignition experiments with firebrand piles likely becomes
less important than the mass of the pile deposited. This dependence on mass may
be critically important in choosing variables to measure during large scale tests or
in the design of standard test methods for materials or components.
Finally, this study confirmed the importance of wind for ignition, but it also
extended that to show how wind dramatically affects the heat flux that a firebrand
pile produces. The ignition of a recipient fuel appears to take place very early after
firebrand piles fall. It is possible that brief flaming from firebrands can act as a
pilot to ignite the recipient fuel as it heats. These results differ from investigative
reports that find that firebrand piles may ignite WUI fuels long after the fire front
has passed. One key difference here is that a full firebrand pile is deposited at once
109
during these experiments, while single firebrands may be added to growing piles
during an actual WUI fire.
8.3 Recommendations for Future Work
There are several areas of future work available based on the results of this
study. One major drawback of the heat flux measured with the thin-skin calorime-
ters was the high uncertainty in the energy balance calculations and the applicability
of the calibration method to this particular experimental set-up. There are two ar-
eas of future work related to measuring heat flux. It should be determined whether
there is a better gauge to provide accurate heat flux measurements. It would be
worth exploring the possibility of using a high temperature thermopile, which would
solve the issue of potential water-cooling. Durability and repeatability would be im-
portant aspects to explore in the implementation of a thermopile. Nonetheless, a
change in heating with area has been shown using the thin-skin calorimeters. It
might be helpful in future tests to provide one good point estimate and a larger
area of cost effective sensors (such as thin-skin calorimeters) around a main sensor.
Additionally, the development of a conduction-based calibration method would
be one potential approach to using thin-skin calorimeters for quantitative measure-
ments. The current calibration was based on an incident radiant heat flux and had
high convective losses. A future calibration might change to a horizontal orientation
and include the larger contribution of conductive heating.
110
This study focused solely on heat fluxes in a flat configuration; however,
crevices and L-shaped configurations are also known to be important in structural
fuels. Measuring heat fluxes during inert tests at different locations in the geom-
etry will be important to understand the influence of these effects. Coupling this
information to ignition conditions for an array of solid fuels typifying the WUI, as
well as variations in brands including different shapes, such as as wafers, and fuels
would help to connect this directly to fire spread and development of standard test
methodologies.
Initial results for a single ignition condition showed higher heat fluxes and
firebrand piles that cooled and then re-heated. Exploring these trends over a wider
range of wind velocities and firebrand pile sizes would be informative for real-world
application and in building a model of the ignition process.
111
Appendix A:
112
The following Appendix includes additional plots. During the text, examples
were given primarily for 12.7 mm diameter firebrands and 9.6 g deposited mass
piles. The following provide additional data on other pile sizes and diameters.
Included are raw thin-skin calorimeter data, raw water-cooled heat flux gauge data,
and comparisons between the thin-skin calorimeters and the water-cooled heat flux
gauge.
A.1 Raw Thin-Skin Calorimeter Results
Figure 1: Heat flux as a function of time measured by the thin-skin calorimeter array
under ambient conditions for 6.35 mm diameter firebrands of 100 g initial mass.
113
Figure 2: Heat flux as a function of time measured by the thin-skin calorimeter array
under ambient conditions for 9.52 mm diameter firebrands of 100 g initial mass.
114
Figure 3: Heat flux as a function of time measured by the thin-skin calorimeter array
under ambient conditions for 12.7 mm diameter firebrands of 100 g initial mass.
115
Figure 4: Heat flux as a function of time measured by the thin-skin calorimeter array
under ambient conditions for 12.7 mm diameter firebrands of 50 g initial mass.
116
Figure 5: Heat flux as a function of time measured by the thin-skin calorimeter array
under ambient conditions for 12.7 mm diameter firebrands of 20 g initial mass.
117
Figure 6: Heat flux as a function of time measured by the thin-skin calorimeter
array under ambient conditions for 12.7 mm diameter firebrands of 1 brand.
118
Figure 7: Heat flux as a function of time measured by the thin-skin calorimeter
array under forced flow conditions for 12.7 mm diameter firebrands of 100 g initial
mass.
119
Figure 8: Heat flux as a function of time measured by the thin-skin calorimeter
array under forced flow conditions for 12.7 mm diameter firebrands of 50 g initial
mass.
A.2 Raw Water-Cooled Heat Flux Gauge Results
The following are raw water-cooled heat flux gauge plots showing all tests,
with averages, and uncertainties.
120
Figure 9: Heat flux as a function of time as measured by the water-cooled heat flux
gauge for 6.35 mm firebrands and 6.3 g deposited mass.
121
Figure 10: Heat flux as a function of time as measured by the water-cooled heat
flux gauge for 9.52 mm firebrands and 8.2 g deposited mass.
122
Figure 11: Heat flux as a function of time as measured by the water-cooled heat
flux gauge for 12.7 mm firebrands and a single brand test.
123
Figure 12: Heat flux as a function of time as measured by the water-cooled heat
flux gauge for 12.7 mm firebrands and 2.7 g deposited mass.
124
Figure 13: Heat flux as a function of time as measured by the water-cooled heat
flux gauge for 12.7 mm firebrands and 5.2 g deposited mass.
A.3 Gauge Comparison Plots
The following are plots comparing the averaged results of the thin-skin calorime-
ters and the water-cooled heat flux gauge.
125
Figure 14: Comparison of average heat flux results using the thin-skin calorimeter
array vs. the water-cooled heat flux gauge. 6.35 mm diameter, 100 g initial mass.
126
Figure 15: Comparison of average heat flux results using the thin-skin calorimeter
array vs. the water-cooled heat flux gauge. 9.52 mm diameter, 100 g initial mass.
127
Figure 16: Comparison of average heat flux results using the thin-skin calorimeter
array vs. the water-cooled heat flux gauge. 12.7 mm diameter, 100 g initial mass.
128
Figure 17: Comparison of average heat flux results using the thin-skin calorimeter
array vs. the water-cooled heat flux gauge. 12.7 mm diameter, 50 g initial mass.
129
Figure 18: Comparison of average heat flux results using the thin-skin calorimeter
array vs. the water-cooled heat flux gauge. 12.7 mm diameter, 20 g initial mass.
130
Figure 19: Comparison of average heat flux results using the thin-skin calorimeter
array vs. the water-cooled heat flux gauge. 12.7 mm diameter, single brand.
131
Bibliography
[1] Manzello, S.L. and S. Suzuki. Exposing Decking Assemblies to ContinuousWind-Driven Firebrand Showers. In: Proceedings of the 11th International FireSafety Science Symposium, pp. 1339-1352, 2014. DOI: 10.3801/IAFSS.FSS.11-1339.
[2] Hadden, R.M., S. Scott, C. Lautenberger, and A.C. Fernandez-Pello. Ignitionof Combustible Fuel Beds by Hot Particles: An Experimental and TheoreticalStudy. Fire Technology, 47:341-355, 2011. DOI: 10.1007/s10694-010-0181.
[3] Caton, S. Laboratory Studies on the Generation of Firebrands from CylindricalWooden Dowels. Masters Thesis. University of Maryland, College Park, 2016.
[4] Radeloff, V.C., R.B. Hammer, S.I. Stewart, J.S. Fried, S.S. Holcomb, and J.F.McKeefry. The Wildland-Urban Interface in the United States. Ecological Ap-plication, 15(3):799-805, 2005. DOI: 10.1890/04-1413.
[5] National Interagency Coordination Center. Wildland Fire Summary and Statis-tics Annual Report 2016. National Interagency Coordination Center, Boise,Idaho, 2016.
[6] Grishin, A.M, A.I. Filkov, E.L. Loboda, V.V. Reyno, A.V. Kozlov, V.T.Kuznetsov, D.P. Kasymov, S.M. Andreyuk, A.I. Ivanov, and N.D. Stolyarchuk.A field experiment on grass fire effects on wooden constructions and peatlayer ignition. International Journal of Wildland Fire, 23:445-449, 2014. DOI:10.1071/WF12069.
[7] Caton, S.E., R.S.P. Hakes, D.J. Gorham, A. Zhou, and M.J. Gollner. Review ofPathways for Building Fire Spread in the Wildland Urban Interface Part I: Ex-posure Conditions. Fire Technology, 53(2):429-473, 2016. DOI: 10.1007/s10694-016-0589-z.
132
[8] Cohen, J. The Wildland-Urban Interface Problem. Forest History Today, pp.20-26, 2008 (Fall).
[9] Theobald, D. and W. Romme. Expansion of the U.S. wildland-urban interface. Landscape and Urban Planning, 83:340-354, 2007. DOI:10.1016/j.landurbplan.2007.06.002.
[10] Cohen, J. Relating flame radiation to home ignition using modeling and exper-imental crown fires. Canadian Journal of Forest Research, 34:1616-1626, 2004.DOI: 10.1139/x04-049.
[11] National Fire Protection Association. The basics of defensible spaceand the home-ignition zone, 2006. URL: http://www.firewise.org/wildfire-preparedness/be-firewise/home-and-landscape/defensible-space.aspx.
[12] Cohen, J.D. and R.D. Stratton. Home Destruction Examination: Grass ValleyFire, Lake Arrowhead, CA. United States Department of Agriculture, R5-TP-026b, 2008 (June).
[13] Quarles, S., P. Leschak, R. Cowger, K. Worley, R. Brown, and C. Iskowitz.Lessons Learned from Waldo canyon: Fire Adapted Communities MitigationAssessment Team Findings. Insurance Institute for Business and Home Safety,2012.
[14] Maranghides, A., D. McNamara, R. Vihnanek, J. Restaino, and C. Leland. ACase Study of a Community Affected by the Waldo Fire – Event Timeline andDefensive Actions. Technical Note (TN 1910). National Institute of Standardsand Technology, 2015. DOI: 10.6028/NIST.TN.1910.
[15] Maranghides, A. and W. Mell. A case study of a community affected by theWitch and Guejito fires. Technical Note (TN 1635). National Institute of Stan-dards and Technology, Gaithersburg, Maryland, 2009.
[16] Maranghides, A., D. McNamara, W. Mell, J. Trook, and B. Toman. A casestudy of a community affected by the Witch and Guejito fires: Report 2 - eval-uating the effects of hazard mitigation actions on structure ignitions. TechnicalNote (TN 1796) National Institute of Standards and Technology, 2013.
[17] Mell, W.E., S.L. Manzello, A. Maranghides, D. Butry, and R.G. Rehm.The wildland-urban interface problem – current approaches and researchneeds. International Journal of Wildland Fire, 19:238-251, 2010. DOI:10.1071/WF07131.
133
[18] Manzello, S.L., T.G. Cleary, J.R. Shields, and J.C. Yang. Ignition of mulch andgrasses by firebrands in wildland-urban interface fires. International Journal ofWildland Fire, 15:427-431, 2006. DOI: 10.1071/WF06031.
[19] Martin, S. Diffusion-Controlled ignition of cellulosic materials by intense ra-diant energy. In: 10th International Symposium on Combustion, pp. 877-896,1965.
[20] Drysdale, D.D. and H.E. Thomas. Flammability of Plastics II: Critical MassFlux at Firepoint. Fire Safety Journal, 14:179-188, 1989.
[21] Rein, G. Smoldering Combustion. Chapter 19. In: SFPE Handbook of FireProtection Engineering, 5th Ed., Ed: M.J. Hurley. Pp. 581-603. Springer, NewYork, 2016.
[22] Anthenien, R.A. and A.C. Fernandez-Pello. A Study of Forward Smolder Igni-tion of Polyurethane Foam. In: 27th International Symposium on Combustion,pp. 2683-2690, 1998.
[23] Gratkowski, M.T., N.A. Dembsey, and C.L. Beyler. Radiant smolder-ing ignition of plywood. Fire Safety Journal, 41:427-442, 2006. DOI:10.1016/j.firesaf.2006.03.006.
[24] Gol’dshleger, U.I., K.V. Pribytkova, and V.V. Barzykin. Ignition of a Con-densed Explosive By A Hot Object of Finite Dimensions. Translated from FizikaGoreniya i Vzryva, 9(1):119-123, 1973.
[25] Zak, C., J. Urban, V. Tran, and C. Fernandez-Pello. Flaming Ignition Behaviorof Hot Steel and Aluminum Spheres Landing in Cellulose Fuel Beds. In: FireSafety Science Proceedings of the 11th International Symposium, pp. 1368-1378, 2014. DOI: 10.3801/IAFSS.FSS.11-1368.
[26] Fernandez-Pello, A.C., C. Lautenberger, D. Rich, C. Zak, J. Urban, R. Hadden,S. Scott, and S. Fereres. Spot fire ignition of natural fuel beds by hot metalparticles, embers, and sparks. Combustion Science and Technology, 187:269-295, 2015. DOI: 10.1080/00102202.2014.973953.
[27] Gray, B.F. Spontaneous Combustion and Self Heating. Chapter 20. In: SFPEHandbook of Fire Protection Engineering, 5th Ed., Ed: M.J. Hurley. Pp. 604-632. Springer, New York, 2016.
[28] Dosanjh, S. and P.J. Pagni. Forced Countercurrent Smoldering Combustion.Proceedings of the Second ASME-JSME Thermal Engineering Joint Confer-ence, 1987.
134
[29] Ellis, P.F.M. Firebrand characteristics of the stringy bark of messmate (Euca-lyptus obliqua) investigated using non-tethered samples. International Journalof Wildland Fire, 22:642-651, 2013. DOI: 10.1017/WF12141.
[30] Santamaria, S., K. Kempna, J.C. Thomas, M. El Houssami, E. Mueller, D.Kasimov, A. Filkov, M.R. Gallagher, N. Skowronski, R. Hadden, and A. Sime-oni. Investigation of structural wood ignition by firebrand accumulation. In:Proceedings of the First International Conference on Structures Safety underFire Blast, 2015.
[31] Dowling, V.P. Ignition of Timber Bridges in Bushfires. Fire Safety Journal,22:145-168, 1994.
[32] Quarles, S.L., L.G. Cool, and F.C. Beall. Performance of Deck Board MaterialsUnder Simulated Wildfire Exposures. In: Proceedings of the Seventh Interna-tional Conference on Woodfiber-Plastic Composites, pp. 89-93, 2003.
[33] Viegas, D.X., M. Almeida, J. Raposo, R. Oliveria, and C.X. Viegas. Ignitionof Mediterranean Fuels Beds by Several Types of Firebrands. Fire Technology,50(1):60-77, 2014. DOI: 10.1007/s10694-012-0267-8.
[34] Manzello, S.L., T.G. Cleary, J.R. Shields, and J.C. Yang. On the ignition of fuelbeds by firebrands. Fire and Materials, 30:77-87, 2006. DOI: 10.1002/fam.901.
[35] Waterman, T.E. and A.N. Takata. Laboratory Study of Ignition of Host Mate-rials by Firebrands. IIT Research Institute, 1969.
[36] Manzello, S.L., S. Park, and T.G. Cleary. Investigation on the ability of glowingfirebrands deposited within crevices to ignite common building materials. FireSafety Journal, 44:894-900, 2009. DOI: 10.1016/j.resaf.2009.05.001.
[37] Manzello, S.L., S. Suzuki, and Y. Hayashi. Enabling the study of struc-ture vulnerabilities to ignition from wind driven firebrand showers: A sum-mary of experimental results. Fire Safety Journal, 54:181-196, 2012. DOI:10.1016/j.resaf.2012.06.012.
[38] Hildalgo, J.P., C. Maluk, A. Cowlard, C. Abecassis-Empis, M. Krajcovic, andJ.L. Torero. A Thin Skin Calorimeter (TSC) for quantifying irradiation duringlarge-scale fire testing. International Journal of Thermal Sciences, 112:383-394,2017. DOI: 10.1016/j.ijthermalsci.2016.10.013.
[39] Manzello, S.L., A. Maranghides, J.R. Shields, W.E. Mell, Y. Hayashi, andD. Nii. Mass and size distribution of firebrands generated from burning Ko-
135
rean pine (Pinus koraiensis) trees. Fire and Materials, 33:21-31, 2009. DOI:10.1002/fam.977.
[40] Forest Products Laboratory. Wood Handbook: Wood as an Engineering mate-rials, pp. 2.4-2.18. United States Department of Agriculture, 2010.
[41] American Society for Testing and Materials. ASTM D4442: Standard TestMethods for Direct Moisture Content Measurement of Wood and Wood-BasedMaterials, 2015.
[42] Miller, C.H., W. Tang, M.A. Finney, S.S. McAllister, J.M. Forthofer,and M.J. Gollner. An investigation of coherent structures in laminarboundary layer flames. Combustion and Flame, 181:123-135, 2017. DOI:10.1016/j.combustflame.2017.03.007.
[43] Safronava, N., R.E. Lyon, S. Crowley, and S.I. Stoliarov. Effect of Mois-ture on Ignition Time of Polymers. Fire Technology, 51:1093-1112, 2015. DOI:10.1007/s10694-014-0434-1.
[44] Singh, A.V. and M.J. Gollner. Local Burning Rates and Heat Flux for ForcedFlow Boundary-Layer Diffusion Flames. AIAA Journal, 54(2):408-418, 2016.DOI: 10.2514/1.J054283.
[45] Muller, E., N. Skowronski, K. Clark, R. Kremens, M. Gallagher, J. Thomas,M. El Houssamia, A. Filkov, B. Butler, J. Hom, W. Mell, and A. Simeoni.An experimental approach to the evaluation of prescribed fire behavior. In:Advances in forest fire research, Imprensa da Universidade de Coimbra, 2014.DOI: 10.14195/978-989-26-0884-64.
[46] American Society for Testing and Materials. ASTM E459: Standard TestMethod for Measuring Heat Transfer Rate Using a Thin-Skin Calorimeter, 2016.
[47] Special Metals Corporation. INCONEL alloy 625, 2013.
[48] Ballestrın, J. Direct Heat-Flux Measurement (MDF) for Solar Central ReceiverEvaluation. Informes Technicos Ciemat, 961, 2001.
136